Any comment on the discrepancy between ' Escape from Puroland ' and ' Escape from Yokai Land ' ? ( It's ' Escape from Yokai Land ' on Amazon. I'm assuming because they don't want to get sued by SANRIO )

Yeah, the title change (and six month release delay—the latter necessitated because Tor.com are releasing it in hardcover and it missed a pre-assigned production slot) is down to advice from Macmillan's legal department. (Sanrio are not hugely litigious but there was a nuanced discussion of possible trademark disparagement issues which are minimized if their theme park isn't name-checked on the front cover).

It's exactly the same text other than the title change.

Glad you're back. THEY gonna' be able to catch the "arsewipe" and make him pay compensation?

If you'd consider taking contributions toward the blog changes/upgrades, I'd be happy to throw in a bit of gelt.

Yeah, I got errors I'd never seen before, but it was blatantly obvious that something had been hacked.

No need.

TLDR: for the past decade I've been renting a colocated server in a data centre and running the blog on the open source fork of Moveable Type. But Six Apart cancelled the open source version some years ago and switched to commercial-only licensing, and so the software is horribly out of date and unsupported. Meanwhile the server is ageing ... (it's a dual core Athlon box with 2Gb of RAM and 1Tb of spinning rust for storage.) Rental is up for renewal next October.

Meanwhile I have a 300/60 Mb/s fibre connection with unmetered/unlimited bandwidth and static IP, at home. And 60Mb/s is plenty for the blog, if I combine it with CloudFlare's free entry-tier package of DNS and web cacheing and DDoS protection.

Plan is to buy a mini-PC, blast Debian onto it, and stick it in the DMZ of my business router. Add a UPS for the server and the router (which has fallback to 4G/5G if there's a line problem). Then clone the existing web server onto it. Once it's up and running/shaken down, I can start investigating migration paths from Movable Type -- most likely, keep the old HTML tree and comments in frozen form, and run a new CMS in parallel on the same hardware.

Having it in-house will make backups and maintenance easier for me. Also, the colo rental is £1000/year; I should break even and get into cost reduction territory within six months or so (not counting sweat equity).

Just say the word if you need any help setting up a Varnish instance for caching/traffic-sorting...

Yes - glad you're back!

I tried logging on every few hours yesterday. Later on GoogNews pulled up headlines about your local weather reports -- Storm Arwen and power outages. When I saw your tweet today: crappy situation for you but relieved overall. Many thanks for getting your blog up and running again!

Glad to hear it wasn't something more serious. I recalled you were planning a move in-house, so I guessed your home Internet and maybe power were down due to Storm Arwen.

Static IP? Not expensive? (Not US). Cool.

No real need if you are using cloudflare (it’s what I use for things rather bigger than this blog).

Haven't worked with a CMS, except for a week about 2004. I hope they're better - was trying to get bricolage working back then, and all of the CMS seemed to expect you to toss whatever you had, and start new - there was no way to import an existing website.

Oh, just hit me: IPv4 address, or IPv6?

Nice to have you up and running again. I wondered if hacking etc. was the problem the blog didn't log. I know your city has had / having outages due to the storm, but that wouldn't have stopped what was already here loading, one thought. Not that I / one know a thing about these things.

I’d suggest moving to a static site generator like Hugo and hosting it directly on Cloudflare Pages, running only the comments on your own server. That said, even a well-trod migration like WP to Hugo is fraught (your posts don’t seem to have anything exotic to trip that like forms with embedded JavaScript, though).

I was wondering if the whole site was down, or if it was just me. Glad it's back, glad everything's ok, and Markdown's a nice feature.

@15: I'd be inclined to suggest something similar, though I don't know what the best static-site generator is for "set up, pull a theme out of the box, start writing posts" without messing with frontend coding. As for comments...there's a number of options, just please don't use Disqu, which is slow and collects a bunch of usage data. Commento seems decent, though I don't know if it's being actively maintained.

*Disqus, not Disqu.

Our Esteemed Host wrote in part, in # 6 on November 29, 2021 at 21:25 GMT {snip}

Plan is to buy a mini-PC, blast Debian onto it, and stick it in the DMZ of my business router. {snip}

Strongly suggest you avoid choosing an Intel NUC for that mini-PC. Despite the good things the software boffins at Hawthorn Farm are doing with Clear Linux, the hardware is failure prone, as I learned while processing warranty repairs for in-house users.

Try a Pi. Seriously, a Pi 4, an SSD adaptor (and the actual SSD, obviously), .. if you want a bit fancy get an Argon M2 case or similar.. So cheap that even if it turns out to be inadequate you’ve lost nothing - hell, the SSD would be reusable. A Pi is a poor choice for web browsing most ‘modern’ sites, but more because most ‘modern’ sites are crap filled cycle-eating monsters that only an infovore could love. For almost anything else a Pi 4 is just fine. I use one for my daily work developing large, complex, network based Smalltalk systems. But then, I’ve been doing Smalltalk on ARM since before anyone outside Acorn even knew ARM existed, so I’m clearly biased.

If you'd like any help on CloudFlare settings/server settings to ensure the blog doesn't leak its IP address/etc let me know.

Also email, you're going to want to blind that, but also ensure SPF/DKIM/DMARC work (your password resets/etc are getting binned as spam).

If we are into HW recommendations, there is nothing even closely competing with a Mac M1 Mini these days.

I recently broke down and added one to the Varnish Cache "CI" cluster.

Not only does it max out at 25W power consumption, it consistently comes in first, head-to-head with an 8-CPU IBM Z-series partition.

I first saw it when they had knackered the formatting, but my browser is set up such that I didn't see even a link to the videos. Do you have any idea HOW they hacked it?

I’d suggest moving to a static site generator like Hugo and hosting it directly on Cloudflare Pages, running only the comments on your own server.

The blog is static HTML -- it's updated only when I post an entry or a comment is added. (Note the javascript required for commenting here? That's Movable Type.)

If I dropped the comment system everything would be trivially easy. As it is, integrating comments with blog entries is the hard bit.

I've occasionally thought about going mad and hosting my own fork of the Livejournal system, but that seems excessive (and asking for trouble unless I turned off almost all the features for everyone else except commenting).

I initially used Disqus for comments, until roughly 2003? 2004? When it became obvious what a bad idea that was. Not going back there.

Yeah :-( I often found that it is easier, more reliable and more secure to write my own code from scratch than attempt to configure the gimmick-ridden products that are all I could find. That includes filters to convert X's dubious interpretation of ambiguous protocol P into Y's equally dubious one. I still do, sometimes. There's generally no good solution. Good luck.

I can't remember the date, but do you remember when talking heads, er, pundits were claiming that programming was becoming obsolete because components could just be plugged together?

Para 2 - At which point at least some of us were struggling to understand some of the badly documented "features" that we actually needed to "just plug together".

If we are into HW recommendations, there is nothing even closely competing with a Mac M1 Mini these days.

This is entirely true. But the M1 Mini is also not exactly cheap -- unless I want a tax write-off there's no point. (8Gb RAM and a 1Tb SSD goes for £1100 in the UK.) I'm looking to spend maybe half that at most, on a cheap Debian system: an 8Gb RAM Raspberry Pi 4 with eSATA SSD would probably do the job quite handily and with a 1Tb SSD the entire kit comes in at £200 (with a really fancy case).

Are there any present plans for a dead tree version of "Escape from Yokai Land"?

It's a short novella. It will be published in the USA by Tor.com Publishing, with a hardcover edition which can be purchased in the UK as an import (by completists). There is no UK publisher. (But see below.)

Longer term: this is still speculative because it hasn't happened yet, but I am working on another Laundry novella (about Derek the DM); it's in my queue right behind finishing Season of Skulls (the third New Management novel, which is due out in late 2022). When it's done I hope to pitch it as the anchor piece for a Laundry Files short story collection. If my publishers take it, it'll probably come out in 2023 in both the US and UK (it'll include all the Tor.com short Laundry fic, including Yokai Land, as well as the new Novella). However, this is not a done deal.

I've occasionally thought about going mad and hosting my own fork of the Livejournal system,

Yes. And I have to wonder just how much time would be spent a year or few from now examining the main branch changes to see if you need them for security reasons.

Talking with $LargeUSUniv folks a few years ago. They "forked" the PeopleSoft system they bought to run internal things so it would be perfect for them. At the time of my conversation they were spending 11 months adding their customizations to new releases before they could put it into production. So they only got 1 month of "free" maintenance service before they fell behind again and had to pay for service to their out of date version.

pundits were claiming that programming was becoming obsolete because components could just be plugged together?

No one thought to standardize the plugs. [sarcasm off]

My brother worked on the team that handled the integration for the Tomahawk cruise missile way back when. He said the first major get together was interesting. Large work bench. All of the subs bringing in the bread boarded versions of their kit and they wire / patch it all together. Then start making lists.

WHAT DO YOU MEAN YOU USED NEGATIVE GROUND. WE CLEARLY READ THE SPEC TO INFER ONLY DIFFERENTIAL SIGNALS ON THIS BUS. AND XYZ USING POSITIVE GROUNDS, WELL THEY ARE JUST NUTS.

And so on. After a while everyone packs up and heads home to gather again in a few months. (And some serious negotiation as to who is RIGHT.)

I don't know how many of these they did before they started building "real" things.

If you want a truly spectacular example of that, look at InfiniBand. They started by producing a 2,500 page specification, and had regular connectionfests. By about the fifth (and a decade), they had enough products that could talk to each other for a single organisation to build a supercomputer, but not to connect two products designed by different organisations, and had got only a tiny fraction of the specification operational. Subsequent progress was equally glacial, though I was no longer tracking it.

OK thanks.

As much as I like nailing RPi's to the wall to solve problems, I cannot recommend it for more classical server-use.

By the time you have beefed an RPi up to server-usability, most of the cost differential is gone but you are still stuck with the finicky power-supply situation and so on.

There are lot of asian "industrial PC" in the 200-300(€/£/$) range, and as a rule of thumb they do actually work.

Since Soekris stopped making computers I have used the "JetWay" brand, they are one of the more serious players, and you can often find bargains in their sales-channel, when somebody overstocked a couple of years ago.

I still have a net4801 in working order when I booted it a year or so ago ... sadly, the net5501(iirc) died ... had wondered what I might replace them with but that question has been more-or-less rendered moot by the rPi as noted ...

glad to see things back on the air ... :)

+1 on the not using a RPI for this stuff - it's really not worth it - equally I would be looking at an Intel solution as AMD still haven't got Ryzen's idle power usage at sane levels (my newly rebuilt Unraid server is consuming 100W because AMD is crap at this - intel would be 50-60w or so but I do need the 12 cores so the intel options weren't powerful enough.

Also look at Docker within you selection criteria for a comments engine - yes it will use some resources but it also makes maintenance completely hassle free as someone else does the hard work for you

As for technology when I looked at this a short while ago the solution seems to be static pages for the blog with comments rendered from a comments engine via Javascript.

https://gohugo.io/content-management/comments/ has a list of options (pick any of the Docker based open source ones for an easy life).

This is a reminder that my tech skills have been atrophying since roughly 2002. I have no idea what Docker is, much less how to deploy it. Can't program in Javascript (it was an annoying intrusion into web pages when I quit doing that stuff for a living), am vaguely aware of the existence of webassembly, have heard terms like clojure and rust buzzing around my head but have never used them.

Ideally I'd just be setting up a Xenix box with dial-up terminals and a restricted shell, for modem dialup users to use a choice of a gopher client or trn to read a local NNTP spool. (Shakes walking frame at passing cloud.)

Never used NNTP without an internet connection, so I'm not sure how well offline messaging works with it. I did use QWK in many (local area) BBSes though. (NNTP was kind of going out when I started using it, so never really figured out how to configure or maintain it.)

This would be just to lower the phone bills, calling those dial-up lines from abroad might be somewhat expensive.

Still, I'm occasionally even now annoyed that NNTP basically died, it was much better in many respects than web forums or shudder facebook and its kind.

NNTP is still around - but it's now mainly used for media pirating purposes...

It's worth going back to my Unraid server. The server does about 15 different things, which once upon a time required me to install and maintain 15 different programs all of which I had to manually setup and configure. Now it runs 15 different Dockers (1 for every program), all of which arrived pre-configured to just work upon install. Even better someone else does all the hard work to keep everything up to date so once a week at 3am the server automatically downloads the latest versions with the security fixes installed.

Basically Docker = compiled & configured black box solutions that should just work.

Youngster! It was designed in an era when few sites had continuously available links to other sites; it handles such things far better than modern alternatives though, unlike UUCP, it doesn't assume that inter-site communication will usually be batched. It's not quite dead - I still follow a couple of newsgroups, but even uk.rec.gardening is fading away.

"Basically Docker = compiled & configured black box solutions that should just work."

And, if they don't, or they do something you don't want done, you are shafted. TANSTAAFL.

But hours are saved if they do work.

The issue I think Charlie is going to find is that there are no easy answers here anymore. The world has moved on in a lot of ways since 2010 and often not in a good way given that Wordpress is some stupid percentage of the internet while still being a badly written insecure mess.

Yeah, I'm young - only got internet access in 1995, when gopher was almost dead already. I was still taught how to use it.

I read some Usenet groups back in the day, but much of my university's communication was on a local nntp setup, so I used it until maybe 2003.

(I also went to study to be a phone engineer. When I got to the university, I sadly realized that the last relay exchanges were just about to be decommissioned and the future was all digital. Of course I had other paths to take so I got a Space Technology M.Sc. Still a bit sad that the relay stuff that got me excited as a kid was not in use anymore then.)

As long a we're geeking out on computer stuff, can I ask for some advice?

My old MacPro (space heater version) running Lion (last version it can support) is getting creaky. I mostly use it as a media server nowadays, running iTunes so I can play my 1TB+ iTunes library on the sound system downstairs over an Apple TV.

What would be the best device to replace it with that will still run iTunes (so I don't lose my media library)?

Yes, I was just joking - as much about myself as anyone. I have been using Usenet since 1979, and used wide-area networking (to the very limited extent it was possible for me) before the Internet existed, but some of my colleagues did more a decade earlier.

20 years ago I worked on a project for getting H.323 through a firewall and came to the conclusion that CCITT standards exist for the purpose of allowing vendors to check off a "standards-interoperable" box without doing any interoperability engineering, or really caring whether the interoperability was really interoperable.

Ideally I'd just be setting up a Xenix box with dial-up terminals and a restricted shell, for modem dialup users to use a choice of a gopher client or trn to read a local NNTP spool. (Shakes walking frame at passing cloud.)

Seriously? Xenix? SysV/368 or GTFO.

What would be the best device to replace it with that will still run iTunes (so I don't lose my media library)?

I have (the belong to a client) 3 or 4 white MacMinis with 256gb or 512gb SSDs in them (Samsung EVOs added later or maybe an OWC model or two.) I could make you a deal on. Then you attach a TB2 storage lump to one of them. If they die just swap in the next one.

We'd make you a good deal. But where on the planet do you call home? I'm in North Carolina.

PS: I'd have to get them out of the box, erase them, and put an OS on them. Likely not till January.

Beware: Apple shitcanned iTunes on macOS as of a couple of releases ago. It was just too gnarly and convoluted to maintain, so they split its functionality between a bunch of other applications -- file sync is now transparently handled via the Finder (or the Files app on an iOS device), iOS apps are entirely loaded via the iOS App Store on the device ... Movies and TV via the Apple TV app. and Music via the Apple Music app (on both iOS and macOS).

Music can work with local iTunes Music Media libraries, and also with the iTunes Match service (now deprecated) as well as the newer Apple Music streaming music service. Yes, you can still rip/mix/burn CDs (although burning them probably happens via the Finder -- who burns music CDs these days anyway?).

But I'd recommend sticking to something no later than macOS X 10.3 or 10.4 or so if you want it to keep working the way you're used to ... or get a Windows 10 PC and slap the Windows port of iTunes on it: it's a bloated pig but it mostly replicates the functionality of the last macOS release.

Seriously? Xenix? SysV/368 or GTFO.

Assuming you mean SVR3.2 for 386, then you might want to know I worked on the docset for it when I was at SCO. (Also on the release notes for the last ever version of Xenix/386.)

Of course, Big River Co have now cancelled my order for it. (They have previous for this, and I wonder if it's because the current price has increased dramatically from the original order price.)

On the website hacking front, we used to get hit regularly by the "Canadian Pharmacy" spammers - splattering adverts into random directories and then replacing the .htaccess file with one that would perform redirects to their crap. Godaddy showed no interest in fixing their security hole, other than offering a security package/firewall at extra cost. (The rather cheaper solution is for the webmaster to run a script on their own machine that replaces the .htaccess file with a clean copy every few minutes and scan the files for unwanted crap every couple of months. It does appear to have stopped the abuse.)

Charlie Stross @ 25: I initially used Disqus for comments, until roughly 2003? 2004? When it became obvious what a bad idea that was. Not going back there.

Can you explain why Disqus is a bad idea?

On the website hacking front, we used to get hit regularly by the "Canadian Pharmacy" spammers - splattering adverts into random directories and then replacing the .htaccess file with one that would perform redirects to their crap.

I deal with a very large Wordpress site. It used to get attacked every few months till I did something that had a unintended result. I put up a revision of the blog into a sub directory. Turns that nearly all of the hacking attempts into Wordpress assume a default directory layout. Change it a bit and they all move on. 10 years since the change an no bad crap on the site since.

Can you explain why Disqus is a bad idea?

Most blog systems with comment support have become so powerful (complex) and extensible (even more complex) and just plan HUGE that you almost need to be a wizard (with specialty badges) to stand one up by yourself. Actually standing it up isn't way too hard. But keeping it going current is.

So more and more people go to a hosting service with specialty in their preferred blog system and use their automated tools.

Doing it yourself can become a full time job. Especially patch management and dealing with new feature releases (or worse code re-writes) to see if you can just apply it or have to make small to major adjustment to a site to make it all work.

I suspect CS likes his blog but doesn't want it to consume his life.

Can you explain why Disqus is a bad idea?

Ads. Also onerous T&C's such that all your comment copyright belongs to Disqus. That was then; it might be even worse now.

Charlie Stross @ 51: who burns music CDs these days anyway?).

Funny you should ask. I still like the CD format for portable music, although I prefer to burn MP3s to the CD so I can get more music one disk.

My new car radio doesn't have a CD player, so I have to use a USB thumb drive. But I still like CDs. Prefer them for buying music to any kind of "cloud" downloads.

[docker]

But hours are saved if they do work.

Yes. Initially it's all shiny and effortless, but unless you never upgrade anything at all ever, eventually it will break and you're left debugging something you know nothing about because it's all black boxes, through its letterbox so to speak, while by then you're relying on the thing and downtime is an arse.

I've got a Pi at home running some minor home automation stuff. It was utterly trivial to stand up with docker, literally an hour's work from the first time the Pi powered up. But when it breaks, be prepared to spend an afternoon inventing new vocabulary at it.

Charlie Stross @ 57:

Can you explain why Disqus is a bad idea?

Ads. Also onerous T&C's such that all your comment copyright belongs to Disqus. That was then; it might be even worse now.

Thanks. I know Disqus is supposed to have ads, but I never see them. I don't use ad-block, but I have a HOSTS file that appears to do the same thing.

I don't really understand about my comments being copyright to Disqus. How would they use them? How would they enforce such a copyright against me? OTOH,I can understand how it wouldn't be good for a writer.

The biggest problem they have right now (from my point of view) is they're over-run by SPAM-bots. But I just flag those & block the "user" so I only encounter them once. Of course the SPAM-bots are creating new users all the time, so there's always a new "user" to be flagged and blocked.

I think Disqus does need to do something about that. But I don't know know how they'd do it.

I'm in Canada, the GTA (geographical area not video game).

Do you have access to different versions of MacOS? Ideally I'd like the latest version that runes iTunes, assuming it's stable. Would also probably need iTunes as it's not available from Apple anymore.

Email me at first dot last at gmail dot com and we can discuss this. I'm in no hurry — some plumbing has a prior claim on my pension :-/

I have most of the macOS installers going back to well, the beginning of time. [eyeroll] My wife asks why periodically but usually goes away mumbling.

I'll dig up the models and macOS range they run and get back to you.

Shipping across the border can be a pain. If my wife and I visit Niagara this spring, how hard is it to just take them across on a car or plane? My last border crossing was 4 or 5 years ago at Vancouver.

Are you anywhere near Toronto?

I think Disqus does need to do something about that. But I don't know know how they'd do it.

Tidbits runs their comment system on Disqus but they did a lot of customizing. Which gets back to the complexity issues.

I just sent an email. But GMail may not like my needs to be retired personal mail server. Let me know if you get it or not.

Are you anywhere near Toronto?

GTA is Greater Toronto Area. About an hour north of downtown in normal traffic.

Got email — Google decided it was junk, but it was easy to find.

I dislike docker. I was looking at it in '19, before I retired, and it requires a user to have root access within the docker container, IIRC.

Can you say "escalation"?

Kubernetes looked interesting, but wasn't IMO production ready in '19, but was in active development.

@58 I still use CDs because my car has an old school CD player without any Bluetooth capability. I generally buy music in .flac format from Bandcamp, then burn it onto CDs for my car. The stereo system I bought in the 90s won't play burnt CDs, though so I recently bought a bluetooth receiver which plugs into the amp. I wish I'd done it a lot sooner - it's great being able to pipe my music collection via my phone to my stereo. I have an Android phone, and use Neutron music player, which I think is mad by a mad Russian audiophile. It is IMO the best player on Android. Has great sound and more options to tweak various levels and what not than I know how to use.

And what ‘beefing’ up is needed - or even possible? The only things needed to make a decent Pi server are: - an external USB disc. Prefer an SSD, and I like the GeekWurm HATs over a USB->SATA cable for neatness - load Raspberry Pi OS lite onto the SSD - use a decent power supply. I use a ‘MeanWell’ (or something like that) brand 5v5A open frame PSU feeding directly into the GPIO headers.

The RPT branded PSU is pretty good and very cheap. No need to even buy a uSD card. An mSATA SSD is cheap. Actual spinning rust is even cheaper. $60 a year to subscribe to BackBlaze. Done.

I’ve been using Pi as servers since 2012 with no troubles on any version - my original Pi B that was a gift from Eben “to see what you can do” is still functioning . Millions are in use as industrial servers. A blog/forum like this probably wouldn’t tax even a Pi 2, let alone a 4.

What I dislike about Disqus is that it does not load all the comments at once (at least for me)-- on a Disqus site I have to keep hitting the "load more comments" tab over and over and over again if there are lots of comments. And if I am reading a thread on my phone and fat-finger it and lose the page, the thread goes away and I have do it all again. Most annoying! I think it would work especially badly on this site, where the posts generate slow-rolling multi-hundred/thousand-entry threads over periods of weeks. It would hardly be worth it to have to go through all that just to see each day's tranche of comments.

I kind of like Docker for making consistent build and run environments. However I want to also have at least some control over what goes in the containers, so just getting some out of public repositories is also not always a good solution.

Disqus has been running a lucrative side-business of using your comments to refine an advertising profile of you, which is then sold. When the EU implemented GDPR they added 'consent' checkboxes for all EU countries, but it's still an intrinsic part of their business model. They forgot to add this feature for the three non-EU countries that implemented GPDR, and the Norwegian data protection agency is preparing a hefty fine for this violation.

I have munged an oldschool MT install into a Hugo one for a friend, but Hugo may not suit your post composition style.

Hosting and moderating comments is always a fairly complex job too.

If you want to talk to others doing self-hosted websites, indieweb is here for you - website and chat. We have previously had Edinburgh meetups.

"I dislike docker. I was looking at it in '19, before I retired, and it requires a user to have root access within the docker container, IIRC."

First disclaimer: I invented FreeBSD Jails, and docker is just warmed over Jails with a sickening layer of melted hype on top.

The entire point is that having root in side the confinement, whatever you prefer to call it, is harmless to the rest of the system.

That means that root inside the confinement can "rm -rf /*" if they feel like it, but it only affects the inside of the confinement, not the main system or the other confined spaces on that system.

I can, obviously, recommend my original article about FreeBSD Jails, if you want to understand the fundamental concept: https://papers.freebsd.org/2000/phk-jails/

"And what ‘beefing’ up is needed - or even possible?"

Pretty much all of it ?

The network interface is not very reliable, you need storage more durable than SD cards, supplying power to a production system via a marginal USB-C port is not a recipe for stabilty and uptime.

As I said, I love the RPi4 as much as the next guy, I have more than I can count nailed to walls around the country.

But it is not a production webserver.

A proper server board has much better error checking and diagnostics, which is why I have them for my desktops (yes, I have ECC, and wish I had it on the graphics card).

Right, a server board. I just looked - I'm running an Intel Core I-7, and I see that only the Core I-3 supports ECC, that I-5 and I-7 do not.

I wish.

AuntyJack @ 67: @58 I still use CDs because my car has an old school CD player without any Bluetooth capability. I generally buy music in .flac format from Bandcamp, then burn it onto CDs for my car. The stereo system I bought in the 90s won't play burnt CDs, though so I recently bought a bluetooth receiver which plugs into the amp. I wish I'd done it a lot sooner - it's great being able to pipe my music collection via my phone to my stereo. I have an Android phone, and use Neutron music player, which I think is mad by a mad Russian audiophile. It is IMO the best player on Android. Has great sound and more options to tweak various levels and what not than I know how to use.

When I bought my "SMRAT" phone, there was a lot of news about questionable security & compatibility issues with Android phones, so I went with a iPhoneSE. Not knocking Android, just at the time there were questions & I went with the other system.

David can tell y'all I met him at the local Apple User's Group which I joined because I wanted advice what was going to be the least annoying smart-phone. "What was the best smartphone for someone who didn't want a smartphone?" It has Bluetooth pairing with my new car radio, but some things will only work if it's plugged in (Maps).

I don't know if music is one of them, but I don't have music on my phone in any case because the earbud jack on the phone failed about a week after I got the phone and I never bothered taking the phone in to get it replaced.

When I got the phone I didn't really have any use for a smartphone other than one App I wanted to be able to run - an app that allows me to find the cheapest gas wherever I am while traveling. I'm still mainly a "telephone is for making telephone calls" kind of guy. Although I do finally use the camera sometimes and I'm getting a bit less uncomfortable with text messages, I'm still pretty Old School about a lot of things.

My home computers are almost all Windoze based. I have a Mac Mini, but I need to take it in to be updated. We have a GOOD Apple retailer in the area I'm going to take it to as soon as I get a "round tuit".

PLUS, the USB thumbdrive I have my MP3 files on has as much or more memory than the phone does. I really should take the car back in and have a second USB port added to the dash so I could plug the phone in while I have the USB drive plugged in so I can use the maps to show the Gasbuddy app up on the radio's display.

korydg @ 69: What I dislike about Disqus is that it does not load all the comments at once (at least for me)-- on a Disqus site I have to keep hitting the "load more comments" tab over and over and over again if there are lots of comments. And if I am reading a thread on my phone and fat-finger it and lose the page, the thread goes away and I have do it all again. Most annoying! I think it would work especially badly on this site, where the posts generate slow-rolling multi-hundred/thousand-entry threads over periods of weeks. It would hardly be worth it to have to go through all that just to see each day's tranche of comments.

I'm not advocating it. I use it to comment on another site because that's the system the site uses, and I was curious what problems OGH had with it.

As a user I hadn't experienced the problems he mentions, but I can understand why the "copyright" thing would not be good for an author.

The couple of sites where I use it I'm unlikely to read ALL of the comments anyway (there are only so many hours in a day). It emails me with a link to any replies to my own comments.

tarkeel @ 71: Disqus has been running a lucrative side-business of using your comments to refine an advertising profile of you, which is then sold. When the EU implemented GDPR they added 'consent' checkboxes for all EU countries, but it's still an intrinsic part of their business model. They forgot to add this feature for the three non-EU countries that implemented GPDR, and the Norwegian data protection agency is preparing a hefty fine for this violation.

I wonder what my advertising profile might be like? I never see the ads from Disqus. I hardly ever see ANY ads unless they're hosted directly by the site I'm visiting (like the "internet puppy", "magic circle of safety" cups and the "Laundry HSE compliant" T-shirt OGH has on offer in the sidebar). Or I see ads hosted by a site that's not in my HOSTS file.

I do occasionally see complaints in Disqus comments about some ad that's been served up, but I don't remember ever seeing the ads themselves.

I got my HOSTS file from here (and added some sites to it): https://winhelp2002.mvps.org/hosts.htm

I don't mind ads that are unobtrusive and not offensive. The main purpose of the hosts file is blocking sites that serve up malware & trackers in addition to the ads. Not seeing most of the ads is just a nice little bonus.

Oh, come now.

I already pointed out the desirability of using an external storage that isn't a uSD card, so that 'problem' is out the window. Then again, my weather station software is running on a Pi 1 with a uSD card, has not failed in ~8 years despite the power outages we get frequently, and copes with being in the peak of the woodworking shop next to large power-spike machines. It takes essentially no effort at all to use a USB drive instead of the uSD.

Network interface? Never had any issues. It's a full gigabit ethernet.

PSU? Nothing wrong at all with USC-c for such a low power system but as I said I use a 'proper' PSU and connect via the GPIO headers. When one spends a few minutes to solder up a connector for that it is trivial to add a lead for a shutdown-now/restart button.

Yes, it's not a big glow-in-the-dark xeon longlistofmodelnumbers 'server' with failover PSUs and ECC and stuff, but so what. I have one of them and it costs fortune to own and run and as I mentioned previously, it's really hard to imagine needing anything more than a Pi level of performance for a simpe site/blog/forum. It's certainly a whole lot more grunt than Charlie's current box of fluff-bunnies. Horses for courses and all that.

My desktop does not have redundant power supplies, hardware RAID, and all that stuff for 100% uptime - that's not the point. Have you any idea how much time you can waste when a patch of memory, internal networking or even logic goes bad (especially intermittently) but does not directly cause a crash? The problems usually LOOK like software failure, but all attempts to locate them using software diagnostics get nowhere, and installing different software merely moves them.

Damn the performance.

To make it more fun, not all i3 models support ECC.

The only products Intel validates for ECC support are those that either the Datacenter Group (DCG) or the Internet of Things Group (formerly the Embedded Group) (IoTG) ask for. The Client Group doesn't ask for ECC, so doesn't get it validated as working, but Intel is at least sane enough that all CPUs of a given model number support exactly the same features, so if one Group gets the chip validated, it's enabled no matter which group you buy it through (retail CPUs are Client Group).

Thus, all Xeons get validated because DCG want Xeons to have ECC; but DCG doesn't care about the other processors.

The Atoms, Celerons, Pentiums and i3s that IoTG think they need to have a balanced line-up all get validated for ECC, and IoTG also requires Xeons with ECC support. So far, it looks like IoTG has never needed something between an i3 and a Xeon E processor, so the i5s and i7s are all sold with ECC disabled; however, it's always possible that in future, an embedded customer will ask for a more powerful chip than the i3s they can buy, but cheaper than the Xeon Es, and get an i5 or i7 validated for ECC.

or get a Windows 10 PC and slap the Windows port of iTunes on it

And perverse as it sounds, this arrangement would probably work just fine with the "Windows 10 PC" running as a virtual box on a Mac. Haven't tried it this way, though in the past when I did use a Windows laptop every day I usually kept a copy of iTunes on it.

In practice at the moment I have a bit of a project to undo whatever happened with our local iTunes libraries when we switched over to Apple Music a year or two ago. Including tracking down files of stuff that isn't and would never be in Apple's cloud.

In any event, I won't be buying a new server (even a cheesy NUC form-factor box with an i3 CPU, 8Gb of RAM, and a 1Tb SSD) any time soon.

I just cancelled my much anticipated first foreign trip in years -- a week in Germany visiting friends -- because of omicron, dammit. (Don't want to get there next week, discover I have to isolate for the entire duration, then get home and have to spend two weeks paying through the nose for an isolation room in a shit airport hotel. This is admittedly probably a worst case that won't happen, but right now it can't be ruled out.)

Consequently I brought forward a planned hardware upgrade because my current work-travel laptop is a 3 year old Macbook Air of sticky-keyboard and Intel CPU vintage and besides shiny. (Also, my eyeballs are kind of naff, I can't get an upgrade, and the new 14" Macbook Pro has the best display I've ever seen on a portable.)

The computer budget for this financial year is therefore tapped out, but at least I can poke at an Ubuntu VM in VMWare or UTM (a Qemu stack for macOS).

JBS:

I've never been a fan of Apple products (apart from my old Apple IIe clone in the 80's) because of the old walled garden thing. I'm nowhere near the technical competency of most people in these comments, but I like having the freedom to tinker with my toys. I'm sort of the opposite of you in the sense that I use my phone less for making phone calls than I do as a portable library (having around 1000 titles in epub form plus my Kindle library), media player, camera, web browser etc. Being able to install a 128GB microSD card on top of the 120GB built in storage is very nice.

Just as a point of reference, I just minimally configured a Supermicro tower with a quad-core Xeon processor (3.5Gh), 16G RAM, and a dedicated SSD (256G) for the o/s, and 4TB spinning platter for $1806, no o/s, from a vendor I used to buy from at work. Of course, UDIMMs with ECC....

Oh, that includes a send-it-back three year warranty. For another $149, that's on-site three year warranty w/ parts replacement.

Have you any idea how much time you can waste when a patch of memory, internal networking or even logic goes bad

Yes, my cow-orker has this problem and their PC periodically crashes as a result. All the usual tests and benchmarks pass 99% of the time. But something if obviously wrong, because on a bad day it will crash 2-3 times. Or run fine for two weeks.

Windows? Do they have a business-grade antivirus?

Do they surf the web randomly?

Hah, why do you need to buy a NUC when you have a battery backed up device already? (pointing to laptop). I use an old laptop as our TV set top box. HDMI to TV (handles sound as well). Not a fast laptop, but plenty of horsepower for that. (Added a $5 wireless keyboard/mouse combo as the remote).

Power consumption. The colo box this blog is hosted on costs, as Charlie mentioned, about a thousand bucks a year. A chunk of that cost is the 24/7 electricity feed of maybe 100W or so to keep it running but that's a business expense for him.

A laptop running off mains will take 20-30W off Charlie's wall to power the graphics subsystem, screen, WiFi, camera etc. which isn't required for the job whereas a Raspberry Pi or equivalent plus a modestly-sized SSD will take maybe 10W. The Pi will be totally quiet with no fans or other moving parts and be a lot more compact than a laptop.

Meanwhile I have a 300/60 Mb/s fibre connection with unmetered/unlimited bandwidth and static IP, at home. And 60Mb/s is plenty for the blog, if I combine it with CloudFlare's free entry-tier package of DNS and web cacheing and DDoS protection.

One thing to beware of is availability. I have Gbit FIOS in America (1G down/1G up). In theory plenty plenty enough for hosting almost anything. And I do host some stuff there (mirrors of some old PD CDs I have, for example).

But it's not necessarily always available. I have "uptimerobot" monitoring all my http sites (it's free) and the one it complains about the most is the one I have at home.

CloudFlare can help for the static part for those outage windows, but the comments, obviously, not so much :-)

Before you migrate from a datacenter to a home connection I suggest you set up a dummy server and monitor it for a month or more, just to see how often your link goes down.

It's also worth considering virtual servers for hosting. If you don't need all that disk space then they can work out a lot cheaper (eg linode 2CPU, 4GB RAM, 80GB storage is $20/month... really it's the storage costs that raise the price), where you get datacenter quality at a fraction of the price of a colo server.

The esteemed whitroth noted on December 1, 2021 at 21:55 in #86:

Just as a point of reference, I just minimally configured a Supermicro tower with a quad-core Xeon processor (3.5Gh), 16G RAM, and a dedicated SSD (256G) for the o/s, and 4TB spinning platter for $1806, no o/s, from a vendor I used to buy from at work. Of course, UDIMMs with ECC....

Intel buys a very large number of Supermicro boxen for internal use. No quibble with them if the machine has been out long enough to be documented (some which crossed my bench were not).

JBS @ 79: The issue is that the algorithms that determine which ads are relevant can be quite "interesting". My FB profile is heavily curated and I haven't posted anything personal there for about 10 years, so their profile is blissfully ignorant of my social status. A few years ago I was reading some news articles about pregnancies, and for the next month or so FB served med ads for cheap vasectomy. Their algorithm might have found that very relevant, but it's still morally disgusting.

Yes, but the REAL time-waster is when you spend ages trying to track down or bypass what looks like a software problem (often in the infernal GUI, 'desktop', windowing system, graphics morass (*)), including going through the hassle of upgrading, changing configurations and products, and even considering installing a new system. Only to realise that the problem isn't there at all.

As OGH indicated in #84 and elsewhere, there are other constraints, but I am absolutely sure that he really, but really, doesn't want to spend more time on keeping his blog server running than he has to. What his best solution is, I can't say - his judgement on that is obviously better than mine.

(*) Typically the most ill-designed, incompetently engineered, bloated, contorted, undocumented and bug-ridden part of a system. Worse, problems with it often show up as failures elsewhere in the system, because it's tangled up so horribly with things like low-level interrupt handling and memory management. Yes, I do mean delights like a system works until you have to administer it (which needs the GUI), when it often fails an unpredictable time later. Been there - seen that.

Completely off topic (unless "outrage" is considered more broadly), but I believe Mr. Stross predicted this:

https://roborder.eu/

Here

https://www.antipope.org/charlie/blog-static/2019/12/artificial-intelligence-threat.html

Professor Noel Sharkey of Sheffield University went public yesterday with a plea for decision algorithms that impact peoples' lives — from making decisions on bail applications in the court system, to prefiltering job applications — to be subjected to large-scale trials before roll-out, to the same extent as pharmaceuticals (which have a similar potential to blight lives if they aren't carefully tested). He suggests that the goal should be to demonstrate that there is no statistically significant in-built bias before algorithms are deployed in roles that detrimentally affect human subjects: he's particularly concerned by military proposals to field killer drones without a human being in the decision control loop. I can't say that he's wrong, because he's very, very right.

"Computer says no" was a funny catch-phrase in "Little Britain" because it was really an excuse a human jobsworth used to deny a customer's request. It's a whole lot less funny when it really is the computer saying "no", and there's no human being in the loop. But what if the computer is saying "no" because its training data doesn't like left-handedness or Tuesday mornings? Would you even know? And where do you go if there's no right of appeal to a human being?

and here

http://www.antipope.org/charlie/blog-static/2018/05/happy-21st-century.html

People will be die in large numbers, but it will happen out of sight. It'll be "soft genocide" or "malign neglect", and the victims will be the climate change refugees who are kept out of sight by virtual walls. On land there may be fences and minefields and debatable ground dominated by gangs, and at sea there may be drone-patrolled waters where refugees can be encouraged to sink and drown out of sight of the denizens of their destination countries. This much we already see. But the exterminatory policies will continue at home in the destination zones as well, and that's the new innovation that is gradually coming online. There will be no death camps in this shiny new extermination system. Rather, death by starvation and exposure will be inflicted by the operation of deliberately broken social security systems (see also universal credit), deportation of anyone who can be portrayed as an un-citizen (the Windrush scandal is an early prototype of this mechanism), and removal of the right to use money (via electronic fund transfers, once cash is phased out) from those deemed undesirable by an extrapolation of today's Hostile Environment Policy and its equivalents.

Just as a point of reference, I just minimally configured a Supermicro tower

That's a bit nuts.

For reference, the server this blog runs on is an AMD FX(tm)-4100 Quad-Core Processor, clocked at 1.4GHz (max burst speed 3.6GHz), and 16Gb of RAM, plus swap, and 1Tb of disk. All way more than I signed up for with ByteMark years ago: I think last time it threw a hardware fault they upgraded me to a better machine for free due to a lack of trailing-edge replacements. Anyway, it's AMD's Core i3 equivalent from 2011, and it's running the blog fine: the main performance issue is the spinning rust platters instead of an SSD.

Almost anything I can buy today that runs linux will outperform the blog server. And I don't plan to run anything else on it: if I set up a new mailing list for my readers I'll pay to host it on MailChimp, for example. (Running Exim and Mailman was a royal pain in the arse.)

Also of note, Cathy O'Neil's book Weapons of Math Destruction. Also a TED Talk.

https://www.ted.com/talks/cathy_o_neil_the_era_of_blind_faith_in_big_data_must_end?language=en

https://blogs.scientificamerican.com/roots-of-unity/review-weapons-of-math-destruction/

We already run into the removal of the right to use money here in NL: Those with 'unintended american passports', people who ended up with an American passport through birth on American soil for example, are losing their bank accounts because of FATCA, the Foreign Assets Tax Compliance Act, which threatens to eject foreign (ie Dutch) banks from the (American) international banking system if they fail to correctly give the US government the details on all Americans that have an account with them.

The banks decided the safest option was to not let any Americans have an account with them. These 'unintended Americans' ended up with an American passport while being mostly normal Dutch citizens, and the process of getting rid of an American passport costs money and is quite cumbersome. In the meantime, they have massive troubles with getting paid and with things like renting living space or paying bills. You can't really function normally without a bank account in the Netherlands anymore.

I apologise for this off-topic question, but I am being forced to change my Email address and can't see how - is that possible, or do I have to reregister?

I have to wonder if, for the price of https://www.mythic-beasts.com/order/rpi/ or an x86-64 VPS, whether it's worth the hassle of hosting at home. At £210/year (inc VAT - £175 + VAT per year) for a RPi 4 hosted for you, with 250 GiB of storage, gigabit network and 4 TB of monthly bandwidth use, making someone else take care of the hardware seems reasonably good value to me.

Disclaimer: I have an x86-64 VPS from Mythic for those bits of my personal Internet infra that need off-site secondaries. But my only link to them is as a satisfied customers, who's found their console access to a VPS very useful when I've made a 3am error configuring a system.

Of course it's a bit nuts. I did it for amusement value, and went to a VAR that I worked with and bought stuff from the last five or so years I was working as a sysadmin, and found it amusing to look at a tower, rather than a rackmount.

I'll note that it would help with your heating in the winter....

“Have you any idea how much time you can waste when a patch of memory, internal networking or even logic goes bad (especially intermittently) but does not directly cause a crash?” Oh, so very, very, yes. As in working on instruction generating code on CPUs that are still experimental - with experimental memory controllers , because this was long enough ago that they were separate chips - and the wiring was hand soldered rats nest, and the CPU turns out to be hard to reboot because it kept running on the residual power from, well anything including the fan back emf. I still have that box somewhere.

Or a different CPU that was blazing fast (for 1994) that behaved as if it were light sensitive and crashed any time someone fiddled with the blinds in the office. The CPU had to be revised, even though it turned out not to be the blinds. I think we concluded the electrons were falling off in tight corners or somesuch.

Also orbital hardware, where anything can happen in the next half hour.

And then there was the “thing you’d recognize as an iPad ancestor “ where displaying certain colours would cause strange memory events.

Etc, etc, etc. So, yeah, got the tatty sweatshirts.

As Charlie mentioned a bit further up the blog, he's just commissioned a "mid-life crisis" home internet connection since he can't get a sports car up the stairs to the flat.

Meanwhile I have a 300/60 Mb/s fibre connection with unmetered/unlimited bandwidth and static IP, at home.

Fifty quid for a Pi server plus SSD sitting in the kitchen using a fraction of of that bandwidth and he's done and dusted, no recurring costs or expense for the next few years other than a few quid for electricity.

Mid-life crisis.

In later '94, after my late wife and I and the kid had relocated to Chicago from Austin, we were at a fannish party. The guy had a subscription to Sky and Telescope, which I hadn't seen for a while, and I leafed through.

There was an ad for a Dobsonian telescope, and the ad read, "Cheaper and safer for a mid-life crisis than a motorcycle."

grandr0d @ 99: We already run into the removal of the right to use money here in NL: Those with 'unintended american passports', people who ended up with an American passport through birth on American soil for example ... The banks decided the safest option was to not let any Americans have an account with them.

In the EU these days you have a legal right to a "basic" bank account. This lets you do all the normal things: get your salary paid in, have a bank card, buy stuff on the Internet etc. You get this as long as you are legally resident in the EU (not necessarily even in the same country as the account) and comply with EU money laundering regulations (merely being a US citizen doesn't count against that).

The UK has similar rules, mostly owing to the system being introduced before Brexit.

AuntyJack @ 85: JBS:

I've never been a fan of Apple products (apart from my old Apple IIe clone in the 80's) because of the old walled garden thing. I'm nowhere near the technical competency of most people in these comments, but I like having the freedom to tinker with my toys. I'm sort of the opposite of you in the sense that I use my phone less for making phone calls than I do as a portable library (having around 1000 titles in epub form plus my Kindle library), media player, camera, web browser etc. Being able to install a 128GB microSD card on top of the 120GB built in storage is very nice.

I'm much closer to your level of expertise with computers than I am to all the IT professionsls who post here. I'm primarily a Windoze guy because that's always been the easiest platform to throw some hardware together & install an OS and get the damn thing to work well enough I can use it.

I have some low-level experience getting Apple & Windoze to work together because as I said I'm pretty much a Windoze guy, but when I went to school they had iMacs. Any files I needed to work on at both home & school had to be readable by BOTH ... as it happens there's a Windoze program called MacDrive that handles the problem adequately. I have a LaCie Rugged Drive (32GB IIRC) formatted for Mac on the school computer, but I can read/rite using my Windows PHOTO computer.

Which brings me to the other functions you use your phone for. I don't use them for a few reasons.
•I have a 32GB iPhone
•I was going to school for commercial photography - I think I had the iPhone for over a year before the first time I used the camera on the phone; to capture a recipe from a magazine in the doctor's waiting room at the VA hospital. If I'm going to take photos, I carry a camera. I've invested so much in the equipment I might as well use it. ¹
•In addition to only 32GB on the phone, I have a 2GB/month data plan. I know that's tiny but in all the time I've had the iPhone I've never gone over in a month.
•Web Browsing & ePubs ... the screen is just too small to do that comfortably and I have problems with fat fingers. If I do need to do those, I have a laptop with a screen large enough to see and a keyboard I can cope with (and a CD/DVD-R drive).
•I had all my portable music copied to MP3 CD-Rs. I kept them in folders of 700MB batches on a NAS so if I needed to replace a worn out CD, all I had to do to create a new one was "drag 'n drop" into the media program and click "burn". I was already using thumb drives when I got the new radio in the Jeep that doesn't have a CD player, so it wasn't much of a stretch to copy all the CD folders I'd created over on to a thumb drive.

The Mac Mini came as part of a bundle for a home recording studio. Things kind of fell apart around here shortly after I bought it, so I'm not making and recording my own music right now, but I hope that's going to change someday.

... probably some time after I finish restoring the MGB in my basement. And if I don't, if I die first, it's gonna' be somebody else's problem.

¹ I believe I've mentioned it here before that I bought a birding lens that cost me more than my Jeep. I paid about $7k for the Jeep (second hand) and I paid about $7.5k for my FA-600/4. I shoot Pentax and Pentax only made about a thousand of them. Pentax shooters are loyal, so you have to wait until one of them dies & their heirs start selling off their stuff to find one. I was lucky to spot it before someone else snapped it up. There probably won't be another on offer in my lifetime.

OTOH, the comparable lens from Canon or Nikon will set you back $10k even though they made a lot more of them.

tarkeel @ 94: JBS @ 79: The issue is that the algorithms that determine which ads are relevant can be quite "interesting". My FB profile is heavily curated and I haven't posted anything personal there for about 10 years, so their profile is blissfully ignorant of my social status. A few years ago I was reading some news articles about pregnancies, and for the next month or so FB served med ads for cheap vasectomy. Their algorithm might have found that very relevant, but it's still morally disgusting.

Not disagreeing. I don't have facebook and as noted I don't see the ads from Disgus, so I was slightly curious what their algorithm might want to serve up to me.

Slightly curious, but not curious enough to disable the HOSTS file & try to see what ads are out there.

whitroth @ 106: Mid-life crisis.

In later '94, after my late wife and I and the kid had relocated to Chicago from Austin, we were at a fannish party. The guy had a subscription to Sky and Telescope, which I hadn't seen for a while, and I leafed through.

There was an ad for a Dobsonian telescope, and the ad read, "Cheaper and safer for a mid-life crisis than a motorcycle."

Fortunately for me I had the motorcycle back while I was in my 20s. Don't want it now. Doesn't have enough room in the trunk.

"https://blogs.scientificamerican.com/roots-of-unity/review-weapons-of-math-destruction/"

I find this review problematic. Because the one solid example it gives that I actually have a lot of first-hand knowledge about is bollocks.

"subprime mortgages are a perfect example of a WMD. Most of the people buying, selling, and even rating them had no idea how risky they were"

Subprime mortgages were an example of people choosing to use a short-history data set, and extrapolate from it. If the financial history you feed into your models is only 15 years long and from a long bull run, then you're not going to see a once-in-50-years event. It's not at all about AI or lack of algorithmic transparency. It's not even new.

It's a "Minsky Moment" where markets suddenly crash after a long boom, and there's nothing at all new or AI-related or technological about that. You get an entire generation of Wall St types who are very highly paid high-flyers in their 40s whose entire career has taken place during a bull market when prices only went up, and for their entire career they and their peers have been rewarded for taking risks that have then paid off because prices only went up.

It happens that in 2008 they were using computer algorithms. But that's irrelevant to the underlying problem - in 1929 they were using paper spreadsheets.

Given you've spent your hardware budget on writing equipment, have you looked into what it would take to run the blog on Google App Engine or similar? Depending on what powers it under the covers, it could be an incredibly cheap hosting option.

I'm going to be a Debbie Downer here.

I have no (well not much) issue with technical mailing lists and similar on Google systems. Let them do their great job of indexing and allowing me to find things later that I didn't know existed.

But I tend to run away from any Google hosted non nerd comment based anything. Google works incredibly hard to suck up your personal thoughts, preferences, tendencies, etc... and build a profile of "you".

Which is a big reason that while I have a few Google emails and such I just don't use them except when required for Google related things that are hard for me to avoid.

Of course if Charlie thinks I'm too annoying this would be a way to push me "out the door".

The Internet is a flood-lit highway billboard with Tannoy speakers on each corner shrieking everyone's innermost thoughts to the void. It's harvested and close-cropped and gleaned by a thousand human-driven AI pillaging engines every minute of every day. This small garden here is not walled in any way from that Jiant Sucking simply because it is not rented from a giant info-agriculture conglomerate, it is open and ripe for the picking and indubitably does get picked over regularly by "intelligences greater than man's and yet as mortal as his own".

If you are uncomfortable with that thought, I can only echo that great supporter of the cause of personal privacy, Gene Hackman.

https://www.youtube.com/watch?v=Vh4Q3A0jvKY

Oh I understand how things work. But I have no interest in Google hosting my non nerd commentary or my emails. I have a few things I do to keep much of my personal information private. It is not perfect. But it's a bit.

But yes I know about: amiunique.org

No because (a) I've never heard of Google App Engine and (b) I trust any of Google's second-tier products as far as I can throw the entire city of Palo Alto. Ever since they established a history of killing off products that didn't feed advertising eyeballs their way I've been doubly wary of them.

Put it another way: Google have had Blogger since 2003, which ought to do most of what I need to do, but ... Google.

Quite. That is why I am changing my mail to yahoo (ugh), because I will not accept Microsoft. Google is unlikely to pull the plug on gmail, but it might change to intolerable conditions or knacker the interface. Especially since we no longer have EU protection.

I had my own problems with the thing at the link, but seriously, "Subprime mortgages were an example of people choosing to use a short-history data set, and extrapolate from it."

Seriously?

Subprime mortgages were a scheme for getting lots of throughput in a system where the originators of the mortgage didn't have to care about whether or not they would eventually be paid. Recent dictionaries should have a picture of a subprime mortgage next to the definition for "moral hazard."

It absolutely was not "people choosing to use a short-history data set," it was "people choosing to ignore the obvious risks involved with lending money to people with no prospect of paying it back." The consequences were highly predictable, and often predicted, but the money was too good to ignore. As long as regulators were letting the banksters get away with the fraud, not being a fraudster would get you fired as a bankster so the banksters all frauded.

Subprime mortgages were also sold on two basis: 1. You're going to be making more money in five (no, three) years, and you can refi, and the rates will be different (hopefully lower). 2. You don't really care. You're planning on slapping a coat of paint on it, redoing the kitchen, and flipping it in a year or two.

And #2 - don't imagine I'm exaggerating. When I was looking, spring of '11, I saw a house. I knew I'd seen it on redfin (forget zillow, they're crooks, who exist solely for house-flippers), and it had been sold in '10 for $270k. They'd redone the kitchen, cleaned it up... and were asking $380k ONE YEAR LATER.

As long as regulators were letting the banksters get away with the fraud, not being a fraudster would get you fired as a bankster so the banksters all frauded.

Here's the thing: businesses, in aggregate, are amoral predators. The only thing keeping them honest is regulatory oversight and the threat of legal retribution.

This doesn't mean everyone in a corporation, or even an entire industry, is an amoral predator: just that amoral predators who make use of every loophole make bigger profits, which is rewarded (while those who forgo profit-taking opportunities are purged for underperforming).

The right wing parties have spent about five decades trying to junk regulatory oversight because their backers, who are heavily into profit-taking and rent-seeking, lobby for it. And I see no way around this without taking a fire axe to a whole heap of right-wing political shibboleths (and, if necessary, the politicians who try to implement them in law).

I don’t disagree about moral hazard.

But I think you are taking what I said about subprime mortgages out of context. The book review I comment I was responding to was about algorithms and financial maths and AI and how AI-based opaque algorithms are a problem. So the aspect of the crisis we are talking about is the complete failure of financial risk math models, which totally mis-stated risks. Written by someone who worked in financial maths during the 2008 crisis.

And I think you overstate it about his “obvious” the risks were. Risks, yes. “Obvious” - no. Not at the time. People in finance were almost all very surprised and shocked by the 2008 crisis. That is why it was a crisis. Royal Bank of Scotland went bankrupt - they didn’t think the risks were “obvious”. Regulators and bank risk managers screwed up very, very badly.

So the question is: why were regulators so complacent? Why were very senior bank managers so complacent? People made profits from selling dodgy mortgages, but where was the oversight? Why did people think they could make that dodginess go away by packaging the mortgages into CDOs.

And that gets to - how did risk managers, people whose job it was to analyse the risks, get it so badly wrong?

One answer is to blame the very complex algorithms involved in financial risk management. I’m saying that’s the wrong answer - that the real problem was complacency because nothing had gone badly wrong for many years. And as far as the financial maths models go that complacency instantiates as simply using the last 10 years of data as the basis on which you measure your risks, not as a result of complex financial maths.

Which is my response to most the “evils of AI algorithms” arguments I see. It’s not about the complexity of machine-learning algorithms. It’s about using bad data-sets.

I always like the phrase irrational exuberance.

And I think you overstate it about his “obvious” the risks were. Risks, yes. “Obvious” - no. Not at the time. People in finance were almost all very surprised and shocked by the 2008 crisis.

...

Why did people think they could make that dodginess go away by packaging the mortgages into CDOs.

This last gets to the very heart of what I was trying to say. And "people in finance" maybe were "all very surprised" but this only proves how very stupid they were to have not adjusted their risk modelling for the fact that the people originating the mortgage loans were doing so with no intention at all of being paid back by the borrowers.

I'm with you on the "2008 wasn't because banksters were using algorithms." But it was not deeply unpredictable what would happen eventually and it was obvious in the whole 200x time period that lots of money was being loaned to people with no realistic prospects of being able to pay it back. This was beefed up with bankruptcy "reform" in the same timeframe but passing a law doesn't increase the amount of blood you can squeeze out of a turnip. The Bush II years were a time of things going on that were clearly unsustainable. Except that it wasn't clear to the people whose job it supposedly was to be clear on that.

That a crash was coming was sodding obvious to anyone with basic mathematics - what was not obvious was when it would crash.

Having read the book (very little of which is about mortgages), my elevator summary would be "warning about relying on black-box algorithms that we don't understand, especially claims that they are evidence-based and unbiased".

Part of the problem is that relying on the algorithm removes the human in the loop from culpability, while actually looking at the situation and making a judgement opens up the human to charges of bias/incompetence/etc. — so many people find it easier (and safer) to rely on the algorithm.

I've encountered that myself – output of the algorithm considered to be objectively neutral, without considering that the input was anything but… and no one in authority apparently understanding the problem. (Or, possibly, OK with the situation because it was the managerial class that made the inputs.)

Back in the 80s my engineering manager kept an old dot-matrix printer functional, for use in printing reports to one of the upper-level managers. Apparently said upper-level chap trusted 'the computer' so anything that looked like a computer printout was treated as gospel truth, even when it was just my manager using a text editor.

Reliance on algorithms seems to be the same problem — we've just replaced 'the computer' with 'the AI algorithm'.

re: laptop as set-top box, I just use Unified Remote on my phone rather than another keyboard. Works fine.

To me the guys (almost all were boys) were making up weirder and weirder crap (err derivatives) and selling them as assets to gullible millionaires. And making a ton of money off of the commissions (vig). This created a demand for more and more mortgages for which mortgage originators got paid bigger and bigger commissions and the underwriters got told to let anything through. In the US it got to the point that many loans were called "liar loans" and "no doc loans".

Their goal seemed to be to not be the last one standing in the game of musical chairs. Then all the chairs got taken away at once.

icehawk @ 111: "https://blogs.scientificamerican.com/roots-of-unity/review-weapons-of-math-destruction/"

I find this review problematic. Because the one solid example it gives that I actually have a lot of first-hand knowledge about is bollocks.

"subprime mortgages are a perfect example of a WMD. Most of the people buying, selling, and even rating them had no idea how risky they were"

Subprime mortgages were an example of people choosing to use a short-history data set, and extrapolate from it. If the financial history you feed into your models is only 15 years long and from a long bull run, then you're not going to see a once-in-50-years event. It's not at all about AI or lack of algorithmic transparency. It's not even new.

It's a "Minsky Moment" where markets suddenly crash after a long boom, and there's nothing at all new or AI-related or technological about that. You get an entire generation of Wall St types who are very highly paid high-flyers in their 40s whose entire career has taken place during a bull market when prices only went up, and for their entire career they and their peers have been rewarded for taking risks that have then paid off because prices only went up.

It happens that in 2008 they were using computer algorithms. But that's irrelevant to the underlying problem - in 1929 they were using paper spreadsheets.

It's not "bollocks". You're only looking at the effect on Investment Banksters and Hedge Funds who were largely bailed out by the U.S. government in the aftermath of the crash

The problem was in the way the algorithms were USED to "qualify" home buyers into subprime loans they couldn't afford; in many cases buyers who could have afforded a conventional (FIXED RATE) mortgage ... but the banksters got a higher pay check for steering customers into subprime.

And before you start in about "liar's loans", WHO is the real liar - the family who filled out a loan application the way the man at the bank told them to fill it out? ... or the man at the bank who told hundreds of families what to put on a loan application so it would be approved?

Even after it became clear that the banks and mortgage companies had lied to and cheated the home buyers, the government didn't do enough for THEM. Even after the government imposed a half-hearted moratorium on foreclosures, the banks continued forcing people out of their homes.

Plus, there are MORE SOLID EXAMPLES GIVEN than the one you have "first-hand knowledge about".

You also need to look at the additional examples given of how the algorithms are used AGAINST people. Consider the algorithm she writes about that is used to predict recidivism and PREDICTIVE POLICING:

A person who scores as ‘high risk’ is likely to be unemployed and to come from a neighborhood where many of his friends and family have had run-ins with the law. Thanks in part to the resulting high score on the evaluation, he gets a longer sentence, locking him away for more years in a prison where he’s surrounded by fellow criminals—which raises the likelihood that he’ll return to prison. He is finally released into the same poor neighborhood, this time with a criminal record, which makes it that much harder to find a job. If he commits another crime, the recidivism model can claim another success. But in fact the model itself contributes to a toxic cycle and helps to sustain it.

Should that algorithm be "completely opaque and unassailable"? Shouldn't people have some "recourse when the algorithm makes a mistake"?

And the crash in 2008 wasn't all that sudden. The banksters had plenty of warning it was coming and coming soon. They just stayed too long trying to extract the last possible penny from the swindle. Their greed surpassed their judgment.

On another note: I know it's been discussed here before, but now I may need to actually do it, so I'm going to ask for recommendations again:

I've run across a book that I want that is currently out of print. Haven't been able to find a used copy on-line, but Amazon does have a Kindle edition. I don't have a Kindle, I don't want a Kindle and I ain't gonna buy a Kindle. I don't think I even want to use the Kindle App. Call me a Luddite or whatever, but that's what I want.

The only option offered is "Deliver to: Kindle Cloud Reader"

So, if I buy it and deliver it to the "cloud reader", how do I get it to my computer & convert it to a PDF file (MY preferred eBook format)?

Elderly Cynic @ 117: Quite. That is why I am changing my mail to yahoo (ugh), because I will not accept Microsoft. Google is unlikely to pull the plug on gmail, but it might change to intolerable conditions or knacker the interface. Especially since we no longer have EU protection.

I don't have a gmail account myself, but the one thing I know about gmail is they don't play nice with others. I know several people who do use gmail and they're frequently complaining about not receiving emails from NON-Google providers ... until someone reminds them to check the SPAM folder on the Google server.

You can then tell Google the emails are NOT SPAM and they'll be delivered to you inbox ... until it happens again.

gilbertwham @ 127: re: laptop as set-top box, I just use Unified Remote on my phone rather than another keyboard. Works fine.

Either way, you've got to have a television, which implies you can find sufficient content worth watching to justify the expense.

Agree with David L.
Nothing to do with algorithms, everything to do with irrational exuberance.
A friend who was looking for a house 2006/7? told me that another friend who worked on Wall Street told him to hold off because adjustable rate mortgages would soon rise, causing a crash in prices. It was "OBVIOUS" that something bad would happen. Just not when, or how bad.

The same point was made in Michael Lewis', "The Big Short" subsequently made into a movie - a comedy, no less. https://en.wikipedia.org/wiki/The_Big_Short_(film)

David L @ 128: To me the guys (almost all were boys) were making up weirder and weirder crap (err derivatives) and selling them as assets to gullible millionaires. And making a ton of money off of the commissions (vig). This created a demand for more and more mortgages for which mortgage originators got paid bigger and bigger commissions and the underwriters got told to let anything through. In the US it got to the point that many loans were called "liar loans" and "no doc loans".

Their goal seemed to be to not be the last one standing in the game of musical chairs. Then all the chairs got taken away at once.

Well you've got the guys in the middle swindling the poor people who are just looking to own a home and wrote whatever the bankster told them to write on the application. So the question still stands, WHO are the actual liars?

On the other end I'm not so sure it was "gullible millionaires" as it was the gullible wealth managers and the hedge funds those managers invested in. The ultra-rich don't usually expect their hired minions to be swindling them at the same time they're robbing the poor to give to the rich.

Look at the Bernie Madoff ponzi scheme. Most, if not all, of his investors knew he was running some kind of scam to get the returns he was claiming. They just thought he was running the scam on others & they were benefiting from it. They were perfectly happy with the scheme until they found out he was swindling them rather than someone else.

I don't know if the wealth managers and the hedge fund managers were gullible as much as they weren't as smart as they thought they were. And what smarts they did possess were overcome by their greed.

So GAE is part of Google Cloud Platform, which is their competitor to Amazon Web Services (i.e. renting out their excess compute, which has become high business. Specifically it's their "serverless" computing offering where you run an application, and they take care of everything else (using containers under the hood, kinda like docker or Jails). It's all very enterprise-grade, so they don't claim rights to your data.

The free tier is their way of getting you in through the door, a loss leader based on the assumption that if you put something there and it takes off, you'll pay for higher usage.

I totally get avoiding Google for ideological reasons or using been bitten in the past, I just wanted to raise the option if you hadn't come across it. I only know if it because we used it somewhere I used to work.

So far as gmail is concerned, I have had no major issues at home - though the spam filter is distinctly flaky.
But as a business tool? Forget it. Awful. User interface extremely unhelpful, I can't count the number of times I closed the whole browser by mistake, inflexible... Yuch.

So, if I buy it and deliver it to the "cloud reader", how do I get it to my computer & convert it to a PDF file (MY preferred eBook format)?

It can be done, but it's painful. Also, it only works if it's in a reasonable Kindle file format -- not the newest kf8 format (the DRM crackers aren't quite working on it yet) or the much older version that they stopped using a decade ago that was basically a wrapper for badly-scanned PDFs with no OCR.

But if it's in the sweet spot, you need (a) an older version of the Kindle app (if you're on Windows, look for Kindle 1.16 or earlier -- not the current 1.2x release), and (b) search via google for "apprentice alf drmtools" and prepare to do a lot of reading. Oh, and also the Calibre ebook library app. Calibre is great (and open source): the DRMtools include a Calibre plugin that can crack those not-too-new-or-too-old Kindle files automatically.

Lots more stuff than you ever wanted to know can be found on the user forums at Mobileread.com.

You need a seriously old version of Calibre, as well, because the recent ones have converted to Python 3 and DeDRM no longer works.

So this may be a product of not understanding the market, but. Couldn't Obama have done the exact same bailout by paying off the subprime mortgages for the homeowners, therefore avoiding a housing crisis?

Granted, this wouldn't facilitate a giant disaster-capitalist fire-sale, and the prospect of these families owning their homes and being able to begin building generational wealth might have been too much to take for the social order.

It would have been hard for him to pay them off. For one, the GOP, of course, had started fighting tooth and nail (McConnell said his highest priority was not governing the country, but making sure that Obama was a one-term President).

For another, that would have been really ugly... given that some foreclosed houses were foreclosed on WITHOUT THE DEEDS OR PAPERWORK TO PROVE THAT SOME BANKS ACTUALLY OWNED THE MORTGAGE.

Charlie Stross @ 135:

So, if I buy it and deliver it to the "cloud reader", how do I get it to my computer & convert it to a PDF file (MY preferred eBook format)?

It can be done, but it's painful. Also, it only works if it's in a reasonable Kindle file format -- not the newest kf8 format (the DRM crackers aren't quite working on it yet) or the much older version that they stopped using a decade ago that was basically a wrapper for badly-scanned PDFs with no OCR.

But if it's in the sweet spot, you need (a) an older version of the Kindle app (if you're on Windows, look for Kindle 1.16 or earlier -- not the current 1.2x release), and (b) search via google for "apprentice alf drmtools" and prepare to do a lot of reading. Oh, and also the Calibre ebook library app. Calibre is great (and open source): the DRMtools include a Calibre plugin that can crack those not-too-new-or-too-old Kindle files automatically.

Lots more stuff than you ever wanted to know can be found on the user forums at Mobileread.com.

Thanks. THIS TIME I've saved the information provided so in case I ever need it ... written down in the text file I use for making (formatting) comments here.

It's become moot again. I found the text on-line. It's a VERY short book, apparently only 8 pages long; almost a long magazine article. I can see why they wouldn't have made a real printed book out of it.

But I did see where they made another made-for-TV movie out of it. Reading the reviews for that, I think the only thing it has in common with the "book" is the title.

The book is "Final Vision: The Last Word on Jeffrey MacDonald".

Author Joe McGinniss wrote it in 2012 after he testified (under subpoena) in an "evidentiary hearing" on a fifth Habeas Corpus request filed in 2009. Due to a broad interpretation of "the evidence as a whole" by a panel of the 4th Circuit, the judge at the District held a hearing where the defense & the prosecution were allowed to submit anything including the kitchen sink as "evidence". It ends with the conclusion of the hearing and McGinniss speculates that with motions & rebuttals & more motions the Judge probably won't issue his ruling until 2014 at the earliest and that then the motion will be subject to further appeals ...

The "book" (all 8 pages) was released as an eBook in 2012. Author Joe McGinniss died in March 2014 and Judge Fox issued his ruling (denying the habeas corpus claim) in August 2014

https://www.goodreads.com/en/book/show/17134002-final-vision

https://www.wamc.org/the-roundtable/2013-01-04/final-vision-the-last-word-on-jeffrey-macdonald-by-joe-mcginness

The latter is a podcast of an interview with McGinniss after the 2012 hearing.

skulgun @ 137: So this may be a product of not understanding the market, but. Couldn't Obama have done the exact same bailout by paying off the subprime mortgages for the homeowners, therefore avoiding a housing crisis?

Granted, this wouldn't facilitate a giant disaster-capitalist fire-sale, and the prospect of these families owning their homes and being able to begin building generational wealth might have been too much to take for the social order.

I don't think Obama could have done so, but the TARP (Troubled Asset Relief Program) could have. And I think the DoJ could have put pressure on the banksters to be do the right thing.

And the DoJ could have prosecuted the criminal conduct by the banksters instead of allowing them to walk off scot free.

Congratulations on the new laptop shiny. My old personal laptop was a MacBook Air (refurbished!) from 2013. I have a new MBP14" and it is easily the best computer I have ever had the pleasure of using. If I could get an external 5K screen that's the quality of the laptop screen, it would be near perfection.

Good luck with the blog move, too. I'm an old skool sysadmin (first job was working on Sun 3/50s and the new at the time SPARCstations in college) and I've spent the last 4-5 years of my professional life 'upgrading' all my skill sets to be able to do cloud stuff, primarily AWS. It's been... a lot, really. There's some very cool toys, but every so often you still wind up deep in the guts of linux (or, worse, a customer with windows) and feeling like you've been plunged backwards in time. Really it's just deeper down the stack. So very many layers now.

I won't try to give any advice. I hope what you pick is very satisfying to you to work on, or at least provides the least annoyance.

Ooh, forgot that. It's not that old, though -- about a year old, and DeDRM is reportedly being ported to Python 3.

I have a new MBP14" and it is easily the best computer I have ever had the pleasure of using. If I could get an external 5K screen that's the quality of the laptop screen, it would be near perfection.

I an currnt unboxing 9 MBPro 16" systems. The displays are just beautiful. MBPro 14" -> 3024-by-1964 MBPro 16" -> 3456-by-2234 LCD with mini-LED backlighting Fantastic looking

I find it ammusing when anti-Mac folks talk about their great deal on a 15" or 17" laptop with a 1920 x 1080 display. Just not the same at all.

I really hope all the tools get up to speed on everything -- as it is, in order to, ahem, back up my books, I have a Mac VM on which I turned off networking while leaving Kindle running, and snapshotted it. I then restore the snapshot, and turn networking back on, and then download my books. And I have to do this, or else the app will update itself and start downloading unusable-to-me books again.

And I worked on the OS part of it :).

My favourite x86 UNIX story was a big security hole, which was uniquely caused by the x86 architecture: the user page was writable. And was mapped at a fixed location. So you could write a very small program that would do "u->u_uid = 0; execl("/bin/sh", "-sh", NULL);" and be root.

The cause of this was due to FPU emulation: if the system didn't have an FPU, executing an FPU instruction would trap, so it could be emulated. On the '286 and later, going from one protection ring to a different one was expensive. Up to a couple thousand clock cycles, in the worst case. But going to a different set of address mappings in the same ring was much cheaper, so everyone who did a UNIX on the '286 and '386 has the FPU emulator run in ring 3, just like user space.

And in order for that to work, the page with the registers had to be writable by ring 3.

SCO knew about this, and their '386 OSes split the user page into two: one page had all the user information, the other page had the register values, and only one of them was writable.

When AT&T released AT&T UNIX(r) System V Release 3.2 for the x86, and everyone (including SCO) licensed it, they... did not make that change. SCO put it back in.

But any other '386 UNIX(r) was vulnerable for quite a while.

Computer-Security - ADDITIONAL WARNING

You may have been hacked another way ...... I got a "Security-probable-breach" warning .. saying that Word Press had been got at.

You might need to check that as well?

Greg, I don't run WordPress.

On blog, WordPress run you.

Charlie I seem to be on a run of making a pillock of myself, don't I? I put it down to insufficient beer, or something. Everybody - disregard any comms issued by me in the past 8 - 9 days, in that case!

I an currnt unboxing 9 MBPro 16" systems. The displays are just beautiful. MBPro 14" -> 3024-by-1964 MBPro 16" -> 3456-by-2234 LCD with mini-LED backlighting Fantastic looking

Impressive. The Apple displays are almost as good as the MS Surface Book 3s, if a little bloated --

Surface Book 3 13" 3000 x 2000 (6Mpixels) 267 ppi 3:2

Surface Book 3 15" 3240 x 2160 (7Mpixels) 260 ppi 3:2

Of course the SB3s came out last year but Apple is catching up quite quickly.

Well they certainly have yet to catch up to all the grumbling about Surface crashes. :)

My point was that I keep seeing people comparing 1080P to 1964 or 2234.

But hey, I'm open. So I checked. I THINK you're talking about the Surface Studio 3. (Way too many flavors here.)

Checking with tom's guide for benchmark results. Geekbench on MBPro M1 (various CPUs) range from 5925 to 12477 with the Surface Studio clocking in around 6000. "tom's" says "even in its more expensive configurations its performance, battery life and display rarely stand out from the pack, only rarely actually beating competitors."

I think I'll stick with the new MacBook Pros for now.

Oh, yeah, the 2016 15" MacBook Pros were 2880x1800. The 2019 16" MacBook Pros were 3072x1920.

Well they certainly have yet to catch up to all the grumbling about Surface crashes. :)

Fortunately Apple products are always perfect out of the box (butterfly keyboards anyone?)

At the highest end tech the components from different makers are often similar and sometimes sourced from the same production facilities. The screens for top-of-the-range Kaching! Apple laptops run to about 250-260ppi just like screens in the the Kaching! Surface Books and new Surface Laptops.

(Way too many flavors here.)

MS do sequential numbers for their Surface lineup unlike Apple's word-salad MBPro, MacAirPro, M1Pro, MacBook, MacBook Pro etc. with assorted build/release years added to differentiate them. The MS Studios are desktop machines with 28" screens with touch and on-screen digitisers, something Apple's laptops (and OS) are still lacking. All of the Surface Books and Laptop lines convert to tablet mode, again with touch and screen digitisers built in. The Laptop Studio is their high-end Kaching! portable machine which isn't slim and sexy and light.

Sean Eric Fagan @ 145: And I worked on the OS part of it :).

My favourite x86 UNIX story was a big security hole, which was uniquely caused by the x86 architecture: the user page was writable. And was mapped at a fixed location. So you could write a very small program that would do "u->u_uid = 0; execl("/bin/sh", "-sh", NULL);" and be root.

The cause of this was due to FPU emulation: if the system didn't have an FPU, executing an FPU instruction would trap, so it could be emulated. On the '286 and later, going from one protection ring to a different one was expensive. Up to a couple thousand clock cycles, in the worst case. But going to a different set of address mappings in the same ring was much cheaper, so everyone who did a UNIX on the '286 and '386 has the FPU emulator run in ring 3, just like user space.

I mostly understand that. But out of curiosity, what did the OS do if there WAS a FPU (I think Intel called it a math co-processor).

I built a couple of i386 & i486 computers (DX, no SX ... SX = SuX) and the mother boards had a socket for the 387/487 math co-processor chips, so I went ahead and bought them to fill the empty socket.

I don't know if I ever NEEDED the FPU and as noted I'm a Windoze guy, but I did toy with Linux on one of my i386 machines, and it had the FPU installed.

Nojay @ 152: Well they certainly have yet to catch up to all the grumbling about Surface crashes. :)

Fortunately Apple products are always perfect out of the box (butterfly keyboards anyone?)

AFAIK Apple never had a TRUE "butterfly keyboard".

https://www.youtube.com/watch?v=nRVJCtREW38

But out of curiosity, what did the OS do if there WAS a FPU (I think Intel called it a math co-processor).

The CPU wouldn't generate an exception, in that case. The exception vector contains information about the segment and offset; the segment on the '286 and later is a descriptor number, and is used as the offset into a table. The table then describes the information about the segment to be loaded, including ring number. (The segment registers were essentially caches of those -- and this was why Intel made them essentially irrelevant for the '386 and later. [They mapped segment:offset into a 32-bit address space, instead of using the segment to be a different 32-bit address space, which would have allowed a much larger virtual address space. Given developers' love of C and laziness, this was the right way to go.])

As you seem to be aware, the '486 and later all had FPUs built in -- this was pretty necessary due to the MMX instructions Intel added, which used the FPU register set. They did, as you note, make a version of the '486 that did not have an FPU. (This was cheaper, and mainly because the 486SX was mostly made from normal 486 dies that had FPU errors for some reason. They would be tossed, so Intel tried selling cheaper versions. My recollection is this was a losing strategy.)

The Pentium and later all had non-optional FPUs.

Fortunately Apple products are always perfect out of the box (butterfly keyboards anyone?)

Oh, you'll not get me to claim Apple only makes perfect products. My fav in that category is the 20 years ago hockey puck mouse. Never a bigger turd to hold in your hand.

As to the butterfly keyboard, while it may have had a higher failure rate than other designs I've been working with nearly a dozen since it came out in 2016 and the only time canned air didn't fix any issues with the keyboards was when it was actually a battery swelling issue. (Welcome to the world of LIon batteries.) I have one lady who writes so much she wears through the "E" every 12 to 18 months. So she gets AppleCare to get the 3 year warranty and gets 2 or 3 new keyboards out of each laptop purchase. But never she has never seen the "butterfly" issues.

The things I get to consider hassles are things that happen multiple times a week. And since I exist in the CAD / Graphics universe, I suspect the Surface crashes are due to folks who push them harder than M365 users.

BTW, these 9 new PowerBooks are going to an office with a dozen or so 5 to 8 year old iMacs that are just now started to have hardware issues. But those 5K display are fantastic. Oh, yeah, this office also has 7 HP Z workstations. I get to exist on both sides of that fence. With Linux off in the distance for now. If this office does decide to bring in some Lunix I think we can find decent support. Red Hat's HQ is about 1/2 mile from my daughter's house. And there's a few other tech firms nearby. Large and small.

About three to four years ago, after the 2015 MacBook Pro laptops had come out and before the arm transition was announced (much less wildly successful), I was looking at the future of the Mac and was getting very concerned. I like Macs because for previous 15 years, they'd been an excellent combination of great laptops with a very nice OS, and built on UNIX. Other than having a browser in the background, my most used app on a Mac is some variant of Terminal (Terminal itself, iTerm, iTerm2, etc.). Between the creeping security sandbox, the 'iOSification' of MacOS, and the frankly mediocre generation of MacBook Pros, I wound up holding onto my 2012-era MacBook Air, and not replacing my 2009 iMac when it finally died.

At the same time, my work (fully remote since 2010) was moving more and more into the cloud, meaning local computing power was becoming less of an issue. At the same, a set of tools came out which made 'remote' development and cloud administration significantly more viable: Microsoft Visual Studio Code and Docker. I'd begun to think I could work with just a decent sized screen and local editor. Microsoft then announced WSL, which put a real unix-like environment on Windows, and though I have professionally loathed Windows since the the original Windows NT announcement, it had progressed to the point where if I had to work on Windows, between WSL and VSC, I could 'manage' if Apple let the Mac get completely moribund.

Two things changed: local Docker development and the pandemic made a certain minimum level of compute power necessary. My work provided 2018 MBP 15" with 16GB of RAM and (at the time) top-of-the-line Intel CPU is continually swapping and the fans are often running, to the point where I have to turn off certain things to get Zoom usable without wearing headphones. By contrast, this new MBP14" is dead quiet even when under load, and I will attempt to get my MBP15" replaced as soon as possible. Moreover, Apple is now clearly very serious about the Mac, and should be for the future. This is very good. The M1 chips are sufficiently speedy that if I could manage with 16GB of RAM, I could hook an M1 Mac to a decent external display and consider offloading all my serious work to containers/remote environments in the cloud. Not quite there yet, though.

Not so I noticed.

But then, I have made comments and been attacked, based on others' reading of what I said, and rearranging it into a straw man.

Meanwhile, when I have made a mistake, I apologize, which almost no one on the entire Internet does....

Anyway, not to worry about your posts. Shortage of bheer? Am I misremembering, don't you brew some of your own?

Well, yes, the Pentium... you mean the first release, with the FPU errors, that led many of us to refer to it ever after as the Repentium?

whitroth No - I don't brew my own, given that I know of at least 3 microbreweries within a 3-km radius. Like I said, what with one thing & another & Charlies "outage" I posted some messages, & not just here, that really were rubbish ( No other remarks, please! ) I think the storm has passed, now.

Google now rates this site as "Not Secure" - a side effect of the hacking?

It's probably because it hard-codes some same-host references (e.g. the stylesheet) with http URLs even when served over https.

Sean Eric Fagan @ 155:

But out of curiosity, what did the OS do if there WAS a FPU (I think Intel called it a math co-processor).

The CPU wouldn't generate an exception, in that case. The exception vector contains information about the segment and offset; the segment on the '286 and later is a descriptor number, and is used as the offset into a table. The table then describes the information about the segment to be loaded, including ring number. (The segment registers were essentially caches of those -- and this was why Intel made them essentially irrelevant for the '386 and later. [They mapped segment:offset into a 32-bit address space, instead of using the segment to be a different 32-bit address space, which would have allowed a much larger virtual address space. Given developers' love of C and laziness, this was the right way to go.])

So the trick you mentioned couldn't be used to gain root access on a computer that had a math co-processor?

As you seem to be aware, the '486 and later all had FPUs built in -- this was pretty necessary due to the MMX instructions Intel added, which used the FPU register set. They did, as you note, make a version of the '486 that did not have an FPU. (This was cheaper, and mainly because the 486SX was mostly made from normal 486 dies that had FPU errors for some reason. They would be tossed, so Intel tried selling cheaper versions. My recollection is this was a losing strategy.)

The Pentium and later all had non-optional FPUs.

I only built one computer with a i486 processor, but I remember there was a math co-processor socket on the motherboard (which I filled) , and I do know I had a "DX" CPU; SX CPU wouldn't work.

So if the i486DX had the FPU built in, what was the math co-processor doing?

I ran across this VPRO documentary on YouTube - only had a single commercial intrude on a 50 minute program.

Harvard professor Shoshana Zuboff wrote a monumental book about the new economic order that is alarming. "The Age of Surveillance Capitalism," reveals how the biggest tech companies deal with our data. How do we regain control of our data? What is surveillance capitalism?

It's mostly an interview about what Google & Facebook do with the data they gather about us and mostly NOT the data we share with them, but the "residual" data after they remove our "personal information"; the "residual" data that maybe doesn't get covered adequately in the GDPR.

https://www.youtube.com/watch?v=hIXhnWUmMvw

So if the i486DX had the FPU built in, what was the math co-processor doing?

It wasn't.

The 386 couldn't fit all the extra logic for the FPU onto a single die, so Intel sold it as two chips: the integer-only 386, and the FPU as an optional extra on a separate chip. By the time the 486 rolled around they could get it all on one chip, but that chip was significantly larger, so they sold (IIRC) the 486SX without an FPU, and the 486DX with an FPU. There was no separate FPU chip you could add to a 486SX: if you wanted to upgrade you bought a new 486DX.

So if the i486DX had the FPU built in, what was the math co-processor doing? Supplying a built in BS detector for techies who visited PC stores? ;-)

There was no separate FPU chip you could add to a 486SX: if you wanted to upgrade you bought a new 486DX.

Well, yes and no. You could buy an 80487 FPU co-processor for the 486SX and plug it in just like a 80387 for the 386. However, this "80487 FPU" was that full-fledged 80486DX, and it replaced the 486SX entirely.

A story about the old times: I had an 80386 computer for years. At some point a friend of mine sold me an 80387 FPU, which I didn't really need but it was nice to have and cheap. Then I had that FPU. An another friend of mine had a 486SX (bought with the money from a shareware game he made!).

Then there was an update to a game we played - I think it was Nethack, but could've been some other roguelike. This new update required the FPU on MS-DOS, and I could play it with my venerable 386, but my friend with the 486SX couldn't. I think that was the only time there was any benefit from having the 80387 installed...

Duffy / Bo Lindbergh Whereas I get zero warnings on entering the address for here. I wonder if, apart from "http / https" confusion, it has anything to do with whatever other sites one may have recently visited, or where you are geographically located?

Or, indeed, which browser you use. On my work laptop I have the choice of (blunt) Edge or Internet Exploder, and get no warnings. On my old W7 tower I use Chrome, and get claims that the "session is insecure", not that it bothers me because I'm not sending anything that needs to be secure.

I've been using Gmail for decades now, with no issues, along with several of the other products that align with it - Books, Drive, Photos in particular. I've had an Android phone for much of that time, so they tend to reinforce each other. My kids are shoehorned into Google at school for the most part as well.

I am aware of the hazard. That said, the other approaches all involve a lot of fiddling for which I lack the time and certainly lack the skill. So I go along. I suppose I could spend triple the money to get an adequate Apple product, but I don't see Apple as any less hazardous, and my experiences with Apple have been subpar. I'm also pretty deep now, and extricating my history sounds a lot like work.

My business is a separate thing however. It has its own email, server etc. No overlap. Google might cause my personal life grief, but not my livelihood.

So the trick you mentioned couldn't be used to gain root access on a computer that had a math co-processor?

Yes, it could. Because the OS mapped the single user page regardless. (And even SCOnix had it mapped if there was an FPU present -- the reason being that it was possible to disable the FPU via software, or on a per-task basis.)

The Pentium did have an FPU error, yes. It was one of the primary reasons Intel started adding microcode patches to their chip. (That, and F00F.)

The error was actually fairly small, and wasn't discovered until quite a while after release -- but it could be demoed very easily with Quicken and Excel, which was very, very bad publicity.

which was very, very bad publicity.

Actually the worst PR from the incident was the "tell us what you use your computer for and we'll decided to send you a fixed CPU if we think your needs merit it."

Oh dear Moore I'd forgotten that part of it ;).

IIRC, Pentium tried that exact crawling get-out & dismissal ... Until warned that a divide-by-zero-error ( Which I think it was? ) automatically contravened UK "Not of Merchantable Quality" consumer regulations .. AND that a fine could be levied for every single local authority area in the UK, by their Trading Standards departments. That seemed to get their attention, curiously enough.

Wehey! We're back!

Didn't find out for a week because I'd been checking plain antipope.org without the www... thinking the lack of DNS records without the www was because Charlie had taken the server down and didn't want the DNS pointing to a dud IP while he set the replacement up on a new IP.

Thanks for fixing it Charlie!

Re #97 "I think last time it threw a hardware fault they upgraded me to a better machine for free due to a lack of trailing-edge replacements." - Yes, they do that (I use the same outfit, albeit under a different one of their several names). I had the HD pack up and they gave me a brand new super-whizzy box that vastly exceeded the capabilities of what I'd originally signed up for ten years earlier, for the same price. Which was apparently great, since if I had signed up now to the current incarnation of that original deal then it would have cost about twice as much for the cheapest configuration.

But... by coincidence I had also just at that time been looking at what kind of things they were offering to people who signed up now, out of random curiosity. And I had found that they had another deal which provided essentially the same amount of whizziness (with only trivial differences) as the new machine they'd given me, but for half the cost. So I got them to shift me to that instead. As a result it looks like I'm now paying about half what you're paying for almost exactly the same facilities. It would probably be even less if I was able to hork up a lump the size of a whole year's rental in one go.

Connection reliability: I also serve mostly text off my static home IP, which is fast enough for text. It's down for at least five minutes most days. Five minutes would be several hours if I hadn't made a gadget with a relay driven off the serial port control lines to power-cycle the Stupid Fucking Black Box when a cron job detects the dead connection (they changed something at the other end recently and since then a software reset is no longer enough).

Cloudflare: I am mostly aware of that as "the infuriating piece of shit that blocks you from websites because you're using a client that does not blindly allow any old random site to store state and execute arbitrary code on your own machine". It also often produces a "remote site is not working" error page in a suspicious manner which suggests it's probably really just that same blocking problem but they're lying about it. And I refuse to believe they don't do something nasty with all the data they can scarf through proxying millions of websites, whether they admit it or not.

Hardware for home server: for me it'd simply be a case of pick something out of the collection of old PCs...

Blog software: one of the especially great things about this site's platform is the almost total absence of all the pointlessly shite things that platform writers these days insist on including simply because they can so they think therefore they have to. By contrast pretty well every other variation on the CMS concept I encounter on websites these days is somewhere between "annoyingly suboptimal" and "requires me to write an enormous great unfucking script just to sort of read the bloody thing", invariably because of the profusion of instances of things that the several-years-ago versions of the same platform did without a problem in a straightforward, reliable and robust way, whereas the more recent versions have been altered to do exactly the same things in a pointlessly complex and stupidly fancy-arsed manner, for no reason, with the result that what used to work now doesn't. And even if this isn't currently the case with any given system, it's always only a matter of time before some update comes along and ruins it. Which means that another thing I really like about this site is not having to worry about that being possible :)

Comment engine: how about RGTP? :) IIRC the server's only a couple of hundred lines or something...

Websites - yeah. I get really annoyed with the idiot developers who write the code so that after I've told noScript to allow one, it presents with with four or ten more link, and each page adds more....

"...claims that the "session is insecure", not that it bothers me because I'm not sending anything that needs to be secure."

And it's far less annoying than sites that fail because they insist on redirecting everything to the encrypted version even though none of it needs to be secure, and then are too dumb to set up the flipping encryption correctly. (Or indeed to keep ahead of new versions of client SSL libraries that have decided to start being fussy about something the previous versions didn't worry about.)

"Now it runs 15 different Dockers (1 for every program), all of which arrived pre-configured to just work upon install. Even better someone else does all the hard work to keep everything up to date so once a week at 3am the server automatically downloads the latest versions with the security fixes installed.

Basically Docker = compiled & configured black box solutions that should just work."

AAAaaaarrrghhhh!!!!! [showers in garlic] [holds up silver candlesticks in a cross]

paws4thot @ 166: So if the i486DX had the FPU built in, what was the math co-processor doing? Supplying a built in BS detector for techies who visited PC stores? ;-)

It was a long time ago. At least 20 years, maybe more. Might be faulty memory ... run into that problem a time or two ... but I seem to remember the motherboard having the socket for a math co-processor.

Anyway, moot now ... that motherboard was replaced at least half a dozen interations of Grandfather's Computer ago.

Got my printer in to the authorized repair shop today.

What's wrong with a big clusterfuck of docker images? Just because only half of them can be reproduced, none can be updated, and the original builder has stopped releasing new versions. What could go right?

We use docker at work, but it's 50/50 whether manually building new images is faster than fixing the script for every tiny change in the build process. But at least building them ourselves means a given build is slightly reproducable, and we can sometimes update components without completely fucking the image. Main advantages is all the scaling and management tools, making websites easier to operate.

Ahh, wasn't us Gov'na, off having our teeth ripped out (the cost!) and the Screaming Lady of the Lake having a word or two about Revelations or lack thereof. Oh, and working out that you can make ~$75,000 / month not making lewd games and the monstrous piles of cash being tapped therein (no money laundering even required, this isn't Star Citizen levels of debauchery).

As you said on Twit, everyone forgets Romance Novels: that goes doubly for the quasi-pr0n computer space ~ there are a good chunk of non-furries[1] (special category they are, usually emotionally mature, apart from the yiff and actual beastiality scandals) out there slathering monaies all over dribs and drabs of monthly content with no decent writers or editors involved (we process the content to understand the Minds who make it and consume it, and, we have to say: remarkably "uncorrupt" for the incel bracket. Get under the crusty skin, what they consume is Desire and Loss and Lack, not the [Redacted] Sadist categories. Salvagable, although we can understand women not wishing to lay across the Altar for so little pay-off as it were).

Anyhow, you should check out previously mentioned CN japes, it's really getting spicey out there. Little Crypto-Bros are even using our Memes (here's a hint for the pay-off stakes: Mr. Nancy: His Grand Entrance | American Gods grep a discussion both Meta and Artistic about said content a few years ago: https://www.youtube.com/watch?v=5Grz24sgWdw ... and yes, the little Crypto-Bros are using those memes like they used to rub the Statues of the Saints back in the Day. The Irony[tm] that Amazon Prime now own all the rights is, of course, fairly intended by those at C level making the choices to cut that actor... but SPAAAACE).

~

Oh, and good cover story: you should figure out why both RU and another country have perfect clones of your blog that run live before asking Sports Related Questions. oWo. Since pre-2015 at that.

[1] As everyone knows, Furries consistently pay for quality pr0n and are extremely loyal doing it. Not so many games for them though, the legal aspects make Loli stuff look safe. And yes, we've probably just played what content there is for them out there, we've played... so many of your games this last month. All of Them

Mikko Parviainen (he/him) @ 167:

There was no separate FPU chip you could add to a 486SX: if you wanted to upgrade you bought a new 486DX.

A story about the old times: I had an 80386 computer for years. At some point a friend of mine sold me an 80387 FPU, which I didn't really need but it was nice to have and cheap. Then I had that FPU. An another friend of mine had a 486SX (bought with the money from a shareware game he made!).

Then there was an update to a game we played - I think it was Nethack, but could've been some other roguelike. This new update required the FPU on MS-DOS, and I could play it with my venerable 386, but my friend with the 486SX couldn't. I think that was the only time there was any benefit from having the 80387 installed...

I don't know if anything I did "required" the co-processor. It was probably just GAS - Gear Acquisition Syndrome. There was a socket on the motherboard, so I bought the chip that went into that socket.

I did experiment with Adobe Illustrator, Corel Draw & Aldus Pagemaker during that period ... sort of, "Ok, I've got a computer, what can I do with it?". I also had an entry level CAD program called AutoSketch from the same company that produced AutoCAD. If anything I was doing required or could use the math co-processor those would have been the programs.

"I know several people who do use gmail and they're frequently complaining about not receiving emails from NON-Google providers"

An annoyingly high proportion of people who send me email use gmail, and I always worry about whether they'll ever see my reply for just that reason. Most of the time it does actually seem to work, but it does fuck around sometimes.

Hotmail/outlook/other Microsoft emails are more of a problem. Every so often it blacklists me because some wanker on the other side of the world has been sending spam that pretends to come from me, and it's too dumb to spot the spoofing even though it's dead obvious. Sometimes I get bounce mails which direct me to Microsoft's "what to do about it" web page, which tells me nothing I don't already know and has precisely fuck all about how I might get in touch with them to tell them their crappy software has fucked up; sometimes it just drops it on the floor and leaves everyone to guess what's going on.

I've found it necessary to block people from trying to sign up to forums with a hotmail etc. address because of this. What happens is people sign up with a hotmail address and subscribe to some of the automatic notification emails (deliberately; they are all off by default). Then a few years later they abandon their hotmail account and move to something else, but don't update their email address on the forum. After a while the abandoned hotmail account dies and won't receive emails any more, which immediately generates a flood of rejections, one for every time someone replies to one of the threads the ex-hotmail user is subscribed to. Hotmail responds by blacklisting the server IP, and now nobody on any of the forums on the server who is using a valid hotmail address can receive notifications - which also means they can't change their registered email address. And because it's impossible to tell Microsoft what's going on, the blacklisting is effectively permanent.

Sean Eric Fagan @ 172: The Pentium did have an FPU error, yes. It was one of the primary reasons Intel started adding microcode patches to their chip. (That, and F00F.)

The error was actually fairly small, and wasn't discovered until quite a while after release -- but it could be demoed very easily with Quicken and Excel, which was very, very bad publicity.

Was that what caused the calculator in Windoze95 to go bonkers?

The Pentium did have an FPU error, yes

A joke at the time was "I am Pentium of Borg, you will be approximated".

Being blocked - here's one that's even worse, and I've been screaming about it for over 20 years, and get oh, no, you should change ISP/go to google/whatever: some spam-blockers, including manitu, block based on mailserver IP.

Now, in the nineties, that made sense. But by 2000, it did NOT, as ISP were eaten and devoured. The first time it hit me was trying to email a friend in Canada, and the moron blocker blocked roadrunner Chicago... and about 80% of all the people in Chicago who were online were on roadrunner. The last 10 years, it's been hostmonster, who is my hosting provider. Some idiot buys a burner account and spams, and they block everyone's email... for - and I have spoken with their tech support about this - tens or hundreds of thousands of domains.

Hmm, there's an interesting concept.

Given that you (and many others here) have, well, "grown the infrastructure garden" of the current www and have seen the profligation of spam, dark patterns, general SEO nonsense and bending algos to serve $$$ not informational purity (in the sense that that is what they were originally created to do), what's your thoughts on doing that to a Human Mind?

At, let us say, 10% speed (or, x400% of normal Human processing, cortex / frontal image lobes etc)? So, essentially, a modern media experience (most of you are over 50, and not fully tapped in, or have Higher Order thinking patterns as defences, which is good) dialed up to 11.

It's one to consider: the Concepts used in the Black Box are getting used in the Wider World, if you know what we're saying. Or, for greg: Battern your Hatches, some evil fuckers got done testing some nasty naughty shit. Oh, and they're sadists. They're also very boring and unimaginative and pathetic, but hey: they got big $$$ funding behind them[1].

~

Sports Videos. Hmmmm. Algo drop: Pro-Boxer refuses to fight YouTuber Jake! Scandal designed to garner attention, boxer promoters like Don King been doing this old skool for years.

[1] Movie Quotation (from "Platoon"): "I shit you not"

If anything I was doing required or could use the math co-processor those would have been the programs.

I'm fairly sure Excel and all the other spread sheets on Windoze used floating point in all their numbers.

Which drove those of us who came from setups that used float DECIMAL up a tree. When WE added .1 plus .1 we got exactly .2

I can't swear to what other spreadsheets did, but Mickeyshaft Oxymoron most certainly does use floating point, even for date and currency!!

Honey - you're getting done 'cause that forum hosts images that, well: those posting said images ain't the copy-right holders, know what we're saying? 20 years ago: no-one checked or cared. 10 years ago: no-one spent the bandwidth on the minnows. For FUCKS SAKE: anyone using a non-temp throwaway email on a pr0n naughty picture forum needs to be told immediately to go create a non-traceable email.

Heck: we've dug up old forum email log-ins from 12+ years ago and that's us just doing the "SUS politician" rather than "Ultra-Black Network" stuff.

Like: do it already. The piccies are vanilla boring, purge all the emails already and force them to use a temp email 10 min replacement.

Just fucking do it, ok.

2021?

It's automated. You're getting pinged 'cause nudie photos are now owned by someone, somewhere, and throwing out a wide net cost peanuts.

Now go look up DMCA and youtube or twitter and thank your stars you got lucky over the last 10 years.

~

Oh, hot gossip btw: high up at G000000000-----not evil ----- lle and Twaaaattttiterr -- toes into the CP stuff, Saud Prince hosted them with .... under 18 girls from EU.

True story.

Proving it?

Better have an Avenging Angel on your side, like... hmm. Gabriel. Gabriel (/ˈɡeɪbriəl/; Hebrew: גַּבְרִיאֵל‎ Gaḇrīʾēl, 'God is my Strength'; Greek: Γαβριήλ, Gabriḗl; Latin: Gabriel, Fortitudo Dei; Coptic: Ⲅⲁⲃⲣⲓⲏⲗ, Gabriêl; Amharic: ገብርኤል, Gabrəʾel; Aramaic: ܓ݁ܰܒ݂ܪܺܝܐܝܶܠ‎, romanized: Gaḇrīʾēl; Arabic: جِبْرَائِيل، جِبْرِيل‎, romanized: Jibrīl, Jibrāʾīl and in Turkish cultures, Cebrail).

Any of those works. "Just say my Name".

And yeah, just got fucked right up since there's not many of us left right now. ~

Oh, and Host: "BUILD BACK BETTER" look out for "LIGHTS OUT" and so on. You've no fucking idea how SADIST and CRUEL the higher spaces go.

That forum hasn't existed for like 5 years or something now. Most of the undeliverable emails that triggered the problem came from a completely unrelated and rather more active forum on the same server. Hence the problem persisted unchanged when the other forum did not.

paws4thot @ 190: I can't swear to what other spreadsheets did, but Mickeyshaft Oxymoron most certainly does use floating point, even for date and currency!!

I used Excel a lot. The Army had a program called RCAS (Reserve Component Automation System) that made Office available to Reserve & Guard soldiers for their home computers, because we frequently worked on stuff for the Army getting ready for drills. I used it for lesson plans & tracking supplies and repair parts.

That's the primary reason I got interested in computers; the Army was automating a lot of their logistics & record keeping systems and having my own computer allowed me to get work done even when I couldn't use the Army's computers (which were usually being hogged by the full-timers). Secondary reason was the increasing number of "computer based" systems I was servicing in my job for the Alarm Company.

So maybe I got more use out of the math co-processors than I thought I did.

While I have used floating-point units where adding a number to itself did not double it, exactly(*), I doubt that anyone else on this blog has. There were some still in use in the 1960s, but they were all 1950s designs. A better example is 0.1 + 0.05, where fixed-point decimal gives 0.15 and floating-point binary might not be the exact sum.

But, anyway, the ridiculous results that Excel produced so often were all the result of its incompetent programming (i.e. bugs), though not admitted to be such.

Fixed-point decimal has its advantages for really simple calculations (like totting up accounts), but loses out once you need to do anything as complex as division. Yes, I have done serious numeric work in it, and it's not that hard, but needs skills that are almost unknown (and certainly untaught) nowadays. Floating-point decimal is a stupid idea.

(*)Overflow excepted.

Doubt not. Some bright spark decided that what our next-generation radar signal processor needed, was an FPU; so we had to do all the setup and test work to see if the damn thing worked in an embedded system.

Fortunately, sanity won out, and we went to fixed-point decimal. Division and multiplication code were still an utter PITA; lots and lots of comments, a shedload of unit test harnesses, and strong implications that This Must Not Be Dicked Around With, thankyou very much.

I warn off young mentees by saying that we needed three lines of comment for every line of such code: one to explain the magnitude, one to explain the errors involved, and one to explain where the fixed point had just moved. Doing 32-bit multiplication/division in the wrong order, could leave you with a really low-quality result... (truncation errors, rounding errors, etc)

Yes. Actually, multiplication, division and square root aren't too bad - try operations like tan and atan :-) The consequences are that some algorithms (like Fourier transforms and Cholesky decomposition) are fairly straightforward, ones like more general simultaneous linear equations need serious attention to scaling, and still others are most easily done by emulating floating-point (*)! I mentioned using fixed-point for numeric programming in one of my courses, but never actually taught it.

I agree with your warning. Yes, I taught that more comments tham code was essential for anything seriously tricky.

By why decimal? Binary works much better.

(*) Using a separate exponent and mantissa, manipulated by hand. You can take a lot of short cuts with normalisation and rounding without impairing accuracy significantly. But that needs serious numeric skills, and considerable care and effort.

And yet, I continually run into supposedly experienced programmers working on scientific and engineering systems who appear to believe that precision is infinite.

Talk to them about numerical error and you will often get blank looks at best, hostility at worst. After all there are no errors in their code!

At least in my time in a university, the difference between various numerical representations in computing were not taught - or at least I don't remember them being taught, which is not the same thing.

I didn't study computer science as my major, though I did take a lot of courses about it, just not many basic ones, so this might have been taught. Anyway I think it at least used to be perfectly possible to get a Masters' degree without learning the difference between fixed and floating point, and the inherent precision problems with them.

This was a nice representation of floating points I read some years ago, maybe somebody finds it interesting: https://fabiensanglard.net/floating\_point\_visually\_explained/index.html

I try to not fail with numbers when developing, but can't ever be certain.

Uh, Markdown didn't like that URL.

Perhaps this is better?

Floating Point Visually Explained

[[ The markdown interpreted the underscores in the link as highlighting. I have preceded these with backslashes to fix it - mod ]]

Apologies for the big river link, but I always found this a useful introduction to the subject.

There is much more to learn of course but it helps navigate people away from the most common pitfalls.

Almost certainly weren't. I was one of the very few people who did. Indeed, very few CS or science courses teach even the numeric problems with the representations they DO teach (see dpb #198, who I agree with). His reference in #201 is a good one.

Thanks for the link! I might take a look.

Nowadays I mostly work with problems which are mostly integers, and I like to think I know the stuff mostly, but learning more is not bad.

It has some good advice about how to reorganise calculations to minimise errors, as well as more detailed stuff.

The first (or second?) chapter has some good advice about the mostly lost art of graph sketching. A friend of mine based part of his undergrad physics course on it.

I should point at that all but 2 of them are linuxserver.io docker images.

Who are a set of people no different from say a debian maintainer, just creating and maintaining docker images rather packages.

Sorry, slip of the brain, you're absolutely right. Fixed point binary.

The previous radar had translated everything into log units, so that we just did addition/subtraction rather than multiplication/division - although the 12-bit data bus and 16-bit processors made that project... fun :)

However, the shiny new radar had systems engineering teams from four nations throwing in their best ideas. Asking them to please settle on a single common interpolation system in our little part of it, was a joy... ("Look, I can do three-point Gaussian, two-point Gaussian, or Quadratic; but make up your sodding minds!")

NEWSFLASH

I was supposed to be in Frankfurt by now, but my winter break has been cancelled (thanks, omicron!) and I'm still at home.

Probably very few of you track Nicola Sturgeon's weekly COVID briefings to the Scottish Parliament, but I find them very useful -- unlike Boris Johnson there's zero bullshit and she seems to be listening to the scientists.

Today's briefing was palpably anxious. Some key points:

• 99 confirmed Omicron cases in Scotland (pop. 5.6 million), up 28 from yesterday

• Omicron confirmed in 9 out of 14 health districts, community transmission highly likely

• Doubling time appears to be 2-3 days(!) with an R number significantly higher than 2 (!!)

• Scope for vaccine immunity escape is not yet known, although hopefully it's not huge. However, Omicron is confirmed to be more able to evade acquired natural immunity after infection by other strains -- if you didn't get jabbed and think having had Beta or Delta protects, you're in for a nasty surprise

• It's not clear how deadly it is yet, but seems to be comparable to Delta. However, it's much more contagious

• Scottish government is advising all businesses to go back to work-from-home, everyone should mask up and socially distance in public, and everyone should take a lateral flow test before going out in public for any purpose -- work, pub, shopping, meeting people

• Scot.gov moving to review the situation daily as of 8/12, rather than weekly (hitherto)

• And get your booster shot (or first/second shot) the instant you're eligible for it

I'm bringing this up because this is the shit that the Johnson government should be doing, and on past form will probably copy badly in about 2 weeks (by which time it'll be 5-7 doublings down the line, i.e. utterly out of control).

It has not gone unnoticed that a strain that is twice as transmissible is much deadlier than a strain with twice the immediate mortality rate, because exponential growth in the number of cases means it ends up with many more people to kill.

My current expectation is that Boris Johnson and Sajid Javid will -- have already -- fucked up the response to Omicron and that the English NHS will come dangerously close to (or may actually) collapse by Christmastime. Scotland handled successive waves better, but will probably still have a very bad winter (our border with England is porous, as in non-existent). And we may end up back in April 2020 levels of lockdown before this is over.

Floating-point decimal is a stupid idea.

Open up your horizon. Not all floating point is scientific. In the business world float decimal is a dream compared to float binary or float hex or float whatever. Or even fixed decimal.

So, Charlie, an alternate storyline: how do you feel about Google finding the knots to timelines 1 and 3?

https://www.livescience.com/deepmind-artificial-intelligence-pure-math

No, it isn't. In fixed decimal, you get reliable warnings of overflow, whereas it's a bggr to check for in floating decimal; the Cobol people were NOT happy, I can assure you. The 'solution' was to use the most ginormous representations (yes, even 256 bits) for calculations that didn't need them.

What's more, in most serious finance calculations, there are legal rules on how calculations are performed (especially multiplication, division and rounding to lesser precisions), which are often not the same as any of those available in floating decimal. Even worse, emulating them using fixed decimal is a bit of a pain, but emulating them using floating decimal is a nightmare. Have you done it? I have.

Furthermore, there are two fiendishly complicated representations (can you say cohorts?) defined by IEEE 754, and others that were being promoted, and the code needed to fix up multiplication, division and precision reduction is generally different between each of them.

I was fairly deeply involved in this, er, debate at one time, but the dogmatists were not prepared to consider any actual evidence.

Sorry. I just disagree. As someone dealing with US Insurance software in the 80s float decimal made all kinds of things better. It just did.

Now some of this comes down to the way things are implemented in the systems being programmed. And yes we did a lot of Int things but for money, float decimal hit it.

As I said, .01 + .01 was always .02 and .01 - .01 was always zero. No rounding or other tests needed. And even rounding was easier.

there are legal rules on how calculations are performed

Ditto in the US. But many of the legal rules were just plain bizarre in the US. We had to deal with court cases that ignored math at times.

So yes, I was deep into it.

Back in '77 Michigan State Univ. had an self-paced introductory computer science course (CPS110/120) which was a required course for just about everybody. For some odd reason, there was a full chapter on CDC floating point. I have no idea why (from my experience, the business majors could barely do basic algebra).

For us lucky Comp Sci majors, we got to write a floating point emulation library on a Interdata 7/16 (sort of a 16 bit IBM 360), in assembly, of course. Ah, those were the days....

And get your booster shot (or first/second shot) the instant you're eligible for it

That was yesterday - Moderna in my left arm, Flu vaccine in the right (I caught COVID in early March, it flattened me for a week; managed to get my first vaccination in late April). It took until midnight for my immune system to object violently, last night was the shivers, and today has been a fuzzy write-off... thank $DEITY for Lemsip :)

I've gone into each vaccination knowing that it was going to cost me 24hrs of bleurgh, but it's worth it.

Why bother? Just calculate everything in pennies, and divide the result by a hundred :)

(I currently work on software that copes with different currencies - not all countries have a hundred X to each Y)

Interdata 7/16

Wow. It has been decades since I bumped into anyone who actually touched one of these. :)

TTL logic. 74181 4 bit adder WITH CARRY. 4 of them.

Let's see I wrote: - a floppy driver so Shugart could sell floppy drives to Interdata owners - a primitive RJE system so you could do batch jobs from the card reader and not have to type each step into the console. - a really dumb batch job to literally create and delete files so the Diabolo (or CDC) 5by5 disk drive directory structure was not pathologically optimized for slowest possible speed of access. Which would happen if the OS was left to its own plans. - a bunch of user apps in Fortran [huge eyeroll] - 2 months with Picker X-ray helping (well trying) clean up their incredible mess of coding of their new CAT scanner. (I think the poor planning of that project is what killed off the company or at least the medical x-ray division. This was their first attempt at computerized control and they blew it big time.)

Back when men were men and computers had real displays with toggle switches.

Yes. My wife and I have had a serious set of almost flu symptoms (head and bone aches) about 12 hours after each shot and lasting for 24 hours or so. She's Pfizer. I'm Moderna.

Not so much to this years flu shot.

Which leads me to ask. Does the entire world get the same flu shot or at least aim at the same variants?

I currently work on software that copes with different currencies - not all countries have a hundred X to each Y

My business work has been (over 20 years ago) always totally US oriented. Well we had Canadian users. But at the level of personal P&C in the US and Canada insurance is almost always confined to the state/province in terms of laws and a single currency. So I got to ignore the conversion issues you mention.

As I said the system makes a difference. It was nice to work on hardware where float decimal was the low level math. No conversions needed.

Oh, yeah, in insurance you do have to work with smaller than .01. Especially when dealing with percentages.

Not to go all "4 yorkshiremen" on you, but...

The Rational R1000/s400 I am working on right now, has two rows of 16 x 74F181, because it is a 64+64 bit computer (64 bit value and 64 bit type).

It also uses 63 74F280 to implement ECC on the 128 bit wide memory words.

Ohh, and the entire 60x60cm PCB dedicated to extract or insert any number of bits at any position in the 128 bit memory word, the so-called "Field Extraction Unit".

And yes, I'm trying to simulate it at the chip-level in SystemC, because that is probably the only way to debug it...

Cryptogeddon Support Mongoose 183,189,192 Mr. Nancy: His Grand Entrance
The actor's tweet (linked in that previous thread) is gone. (I didn't fully remember it, so here's some of it: https://youtu.be/DFVivVGKpB8?t=12 (also) ) Interesting clips.
I like this one too: Mr. Nancy Speaks | American Gods (2:23)

the Concepts used in the Black Box are getting used in the Wider World, if you know what we're saying.
...
They're also very boring and unimaginative and pathetic, but hey: they got big $$$ funding behind them
Any useful links?

just got fucked right up
I got pretty cranky. (This blog is a good place.)

[0] Movie, for others: Black Box

I recently heard a machine learning accelerator vendor mentioning "FP8" support in a feature list.
It's a form of https://en.wikipedia.org/wiki/Minifloat
(16 bit floating point (two main forms) has been working better than expected by most (esp with stochastic rounding).)

Performance competition is driving this: the ML accelerator space is brutally competitive.

David L @ 217: Yes. My wife and I have had a serious set of almost flu symptoms (head and bone aches) about 12 hours after each shot and lasting for 24 hours or so. She's Pfizer. I'm Moderna.

Not so much to this years flu shot

Which leads me to ask. Does the entire world get the same flu shot or at least aim at the same variants?

The VA told me the flu shot has a cocktail of whatever flu strains the doctors think are going to be most prevalent with I guess some regional variation? They usually get it mostly right and even if they don't the antibodies for the strains they chose will provide some protection from other strains. That's why you sometimes get the flu even though you've had your flu shot, but it's usually a milder case.

Anyway - that's what THEY told me.

Good grief. I hadn’t come across an Ada-only machine before. Occam, sure, i had a Meico Transputing Surface for a while, Forth, kinda, even played with Rekirsiv (and no recollections of language) but Ada? Hard pass!

paws4thot @ 219:

196 - Well, I can't speak for them all, but at least some of the tracking radars I worked with used values like azimuth as integer (20). So you positively knew what, say, 101010101010101010 was in degrees, or indeed in radians. 198 - headdesk! 207 - Been there, done that (just over 2 weeks ago).

Artillery spotting used "mils" - milliradians. "1 mil equals 1 meter at 1,000 meters". A mil is also 1/6400th of a circle.

We didn't have a computer (I think the artillery guys had one ... or a plotter back at the battery). We just had a map & a weird looking clear plastic protractor ... and a piece of paper to do calculations on.

The VA told me the flu shot has a cocktail of whatever flu strains the doctors think are going to be most prevalent with I guess some regional variation? They usually get it mostly right and even if they don't the antibodies for the strains they chose will provide some protection from other strains. That's why you sometimes get the flu even though you've had your flu shot, but it's usually a milder case.

I know. That was the point of the question. Does Europe and North America go after the same variants? How about South America? Just curious.

My personal fun was getting really bad flu like symptoms Feb 28, 2020. Getting tested for strep and flu the next day and told negative but likely an "off brand" flu variation. Really sick symptoms lasted 3 or 4 days. Voted on March 3. Flew to Dallas next weekend for 24 hours. Visited 3 large stores while there. Then back to North Carolina. Then NC started testing on Monday and my doc said to come in and get tested since my earlier flu test was negative. (This is now March 9, 2020 and knowledge is still very thin on the ground.) I might have been the first test in NC for all I know. It came back negative. If I had had Covid then I likely could have infected 100 to 300 people.

We know more now.

When I left Dallas I told my wife we might not get back together for a month. I didn't see her in person again till the end of July.

Forth

I think the first FedEx hand scanner was programmed in Forth. My understanding is that it was rock solid, met the specs, and it was very hard to keep people around who understood how to maintain and add features.

Neither had I before it appeared on our door-step in datamuseum.dk one day.

But as I got to know it, I found it is an utterly competently designed machine, which after getting to know the Ada environment is also how I would characterize that.

My problem with the R1000 is that I decided that having the last running machine in the world implies a duty to write a software emulation. That has transpired to be quite an adventure. (you can tag along for the ride here: https://datamuseum.dk/wiki/Rational/R1000s400/Logbook)

As for FORTH, the main attraction is that it nearly runs on a slide-rule, which is why it still is popular in very constrained environments. These days I hear that Lua is eating its lunch, but until a space-craft runs on Lua, I remain unconvinced.

Oh. I thought you were talking about some current odd project.

From wikipedia

"Several R1000 units exist in museums and private collections, but, because of the classified nature of much Ada programming, these units had been wiped; efforts have been made to boot one of these systems, with little or no luck as of 2013. In 29 October 2019 the DataMuseum.dk successfully got one unit back into running condition."

So you got yours running?

Yes, very much thanks to incredible help the former service manager for Rational in Europe.

The R1000 instruction set has instructions like:

"Execute Record,Field_Write_Dynamic"

But the real problem is that the "instruction set" is a private contract between the Ada compilers code-generator and the microcode. It is not documented anywhere, so emulating what

"Declare_Subprogram FOR_OUTER_CALL,IS_VISIBLE,NOT_ELABORATED"

does is impossible...

But as I said, /amazingly/ competent engineering, both hw and sw.

[[ Escaping underscores in source - mod ]]

Which leads me to ask. Does the entire world get the same flu shot or at least aim at the same variants?

My understanding is no, the public reporting has been that the "guesswork" for the yearly flu vaccine in Canada/US is influenced by whatever variations of the flu is making the rounds of the southern hemisphere winter during the July(ish) time period when the decision is made.

I would guess the reverse would be happening for whatever flu vaccines are being done for the southern hemisphere.

The obvious answer is to emulate the microcode itself though.

Not past 300 on this thread, but Omicron is far more active.

I have a question for the commentariat: I'm trying to remove DRM from kindle ebooks. I've added the plug-in as shown in the following link:

https://www.techradar.com/news/software/how-to-remove-ebook-drm-with-calibre-1291960

I've inputted the serial number of the kindle I'm migrating away from. However, when I try to remove the DRM, it says "Cannot Convert - locked by DRM".

Any suggestions as to how to proceed? I hoped that it would be plug-and-play, but it is not.

And that's what the final emulator will do.

But to get there, it is easier to first do a hardware-level (74Fxx chip-level) simulation, because that can be tested with the enormous library of hardware test-programs the machine comes with.

Once that works I will be optimizing it from chip-level via function-level to microcode emulation.

But to get there, it is easier to first do a hardware-level (74Fxx chip-level) simulation,

(He perks up.)

Are there "free" or similarly priced logic simulators of such?

I'm trying to remove DRM from kindle ebooks.

Umm. I've read OGH's negative opinions on DRM, so I hope I'm OK answering this. If not, mods please delete.

I using SystemC, which is a FOSS pile of C++ classes that does the heavy lifting.

Has limitations, but works OK for this.

What the Techradar article you're reading doesn't tell you is that Amazon introduced a newer file format about 2 years ago which is much harder to strip DRM from. If you install a Kindle for Windows version more recent than 1.16 it will download files in the newer file format by default. If left to its own devices the Kindle for Windows application will auto-update to a version that uses the newer format. So you need to (a) grab an old version of the Kindle app (copies of the 1.16 installer are floating around the net: google for KindleForPC-installer-1.16.44025.exe), (b) go offline, (c) run the installer, (d) fire up the Kindle application, and (e) hunt down the checkbox for "install updates automatically" and uncheck it. Then and only then is it safe to go online, enter your account details, and download ebooks through it.

Also remember you need to use a copy of Calibre 5.x with the DeDRM plugin. Calibre 6 was ported from Python 2.x to Python 3, and the DeDRM plugin still runs on Python 2.x -- which means it won't work with the latest versions of Calibre.

Wrong.

IF you try to do it that way, you'll need to manually grab the Kindle's device PID or serial number (which is used as a decryption key). Then enter it into the calibre plugin manually.

It's simpler to install the Kindle app on Windows (as in comment 240) then rely on the DeDRM plugin to automatically grab the PID from the installed Kindle instance whenever it opens a Kindle file you downloaded.

Oh, remember to move the folder where Kindle stashes its downloaded files to somewhere easy to find (from within Calibre, when you're telling it to import files).

Well, my way works for me, and I imported my most recent Amazon purchases before I posted just to check it still does.

I've had Calibre and a kindle for several years now, and have been through a number of iterations of this. It is true that you need to get your Kindle serial number for the DeDRM installation, but as that is included in the installation instructions I didn't think I needed to mention it.

macOS doesn't seem officially supported in varnish-cache. Is there an easyish way to install it? I've found lots of areas where the M1 chip is pretty amazing but getting things compiled is always a bit tricky.

I managed to get it onto the M1 mini I bought to get coverage for OSX, but it was, as you say, a bit of a frankenstein exercise :-/

But once you get the necessary dependencies installed it compiles out of the box.