Some of you probably know about Scrivener, the writer's tool from Literature and Latte. (If you don't, the short explanation is that it isn't a word processor, it's an integrated development environment for books. It's cross-platform (although initially developed for Mac OS X —versions for Windows and Linux are available, and it's being ported to iOS and Android), modestly priced, and has more features than you can wave a bundle of sticks at, mostly oriented around managing, tagging, editing, and reorganizing collections of information including rich text files.) I've used it before on several novels, notably ones where the plot got so gnarly and tangled up that I badly needed a tool for refactoring plot strands, but the novel I've finished, "Neptune's Brood", is the first one that was written from start to finish in Scrivener, because I have a long-standing prejudice against entrusting all my data to a proprietary application, however good it might be. That Scrivener was good enough to drag me reluctantly in is probably newsworthy in and of itself.
First of all, I should note what Scrivener can't do for an author.
Many publishers these days have moved to electronic document workflow during production. Manuscripts are submitted in a standard format (they've settled on the hideous, proprietary, obsolete binary format of the Microsoft Word 97-2003 .doc file, simply because that's what most people use). Copy edits are applied to the .doc file using Word's change tracking feature with annotations in place of post-it notes. If you want to process copy edits in this brave new world, you need a word processor, because Scrivener's view of a book is so radically different from Microsoft Word's single monolithic file that there's no way to reconcile the two and add Word-style change tracking to Scrivener. Luckily LibreOffice, a free fork of OpenOffice, is (a) free, (b) under active development again, and (c) can chow down on basic Word documents with change tracking and notes without throwing up most of the time. (The copy-edited manuscript of a novel does not contain Word BASIC macros, complex tables, or illustrations: it's just a stream of text with paragraph styles.)
So Scrivener stops supporting publisher workflow once you have submitted the manuscript. And arguably it stops an hour before then, because figuring out how to modify the output format generated by the Scrivener "Compile" menu option is a black art ... I found it easier to slurp the resulting Word document into LibreOffice for final tidying up and reformatting before I submitted it. Scrivener doesn't support Word's paragraph style mechanism as far as I can tell; it simply emits styled text. So it's output isn't a direct product you can feed into an unattended turnkey pre-press package: you'll still have to pay someone to drive InDesign for you.
Other weaknesses: Scrivener 2.3 on OSX is a big program. There's an introductory tutorial project, and a video. And then there's a 300+ page manual in PDF. Why PDF, when Scrivener emits some of the cleanest epub files I've ever seen? And why doesn't it work with the OSX built-in help system? Who knows. Let's just say that learning Scrivener's ins and outs is an ongoing task. (For example, I was most of the way through this novel, the (counts) sixth that I've used Scrivener for to some extent, when I discovered that the Edit->Writing Tools submenu now contains a character name generator, as well as the obvious stuff like the spelling and grammar checker controls.)
In Scrivener, if you're writing a book you start by creating a new project, just as you would if you were starting to write a program using an IDE like XCode. The project is a hierarchical outline-based container for your research notes (including PDFs and images and web pages, which you can slurp in as files or direct from the web by entering URLs) and the small files, or "scrivenings", that constitute the work in progress. Scrivenings are basically RTF files (more accurately, Apple's RTFD—a derivative format that allows the inclusion of additional sub-elements like images), or folders containing scrivenings. A chapter is basically a folder, and the scenes in the chapter are scrivenings, and you get a collapsible, hierarchical view. You also get the ability to edit scrivenings, either individually, or by multi-selecting a bunch of them and seeing them as a continuous scroll of text: most convenient if you want to edit scenes 1, 2, 4, 6, and 8 in a chapter but not 3, 5, and 7, for example.
That's treating it as a scene-based word processor. Scrivener provides other tools for looking at your data. There's a cork-board, in which you see each scrivening as an index card, and in which metadata (notes, defined keywords, all sorts of stuff) is transparently visible. Or you can display it as an outline in a classical outline processor mode. The general effect is to make it easy to search, organize, and see views of your data, and trivially easy to restructure a hierarchical document as long as you've broken it down properly into chapters containing sub-documents.
Other tools: in some ways the most useful feature it provides for a jobbing author is the Project->Show Project Targets option. You get a floating window with progress bars (updated in real time) containing (a) your progress towards the target word count for the entire document, and (b) your progress towards your target word count for the day. As motivational goads go, this one is invaluable when you're slogging through the difficult middle of a book, and the ending seems as far away as the beginning. (Seriously, measuring your progress is one of the under-stated but vital tasks associated with any job: good luck getting Microsoft Word to help you with that.) Again: Scrivener projects can get quite large, and are structured internally as a folder hierarchy. Scrivener has an option to package them up as a zip archive (which can be emailed around, or re-imported later), and also to back them up to a private folder. Mine is linked to my (private) Dropbox account, for obvious reasons: it gives me version-controlled offsite backups. It's not quite git or subversion, but if you want those, there's a "sync with external folder" option which looks like, yes, you could use it to sync with a heavyweight configuration management system. (Note: in my opinion, novels don't need heavyweight version control—they virtually never fork and you seldom have as many as two authors. Straight linear versioning is fine for 95% of cases.)
Stuff I don't use: there's a full-screen mode for folks who like to write without distractions. They are not me, and I just don't use it. The keyword tagging ... I can see types of work it would be useful for, but it's less obviously useful for fiction. Being able to define the status of a scrivening as planned, first-draft, or final is obviously useful to some people: but that's not how I work.
Finally, there's the question of how you get your data out of the application. You can do it piecemeal: Scrivener is happy to export individual scrivenings or files. Or you can do it wholesale, via the File->Compile menu. Which takes the assembled scrivenings, filters them in accordance with whatever crazy criteria you set ("exclude odd-numbered scrivenings in even-numbered chapters" looks like it ought to be possible), applies transformations to them (Scrivener understands MultiMarkDown, so if the idea of proprietary RTF brings you out in cooties you can write in MMD text files), and generates a finished document in one of the target output formats—Word .doc is one, but it can also produce RTF, PDF, ODT, Final Draft, and ebook formats—epub or Mobi for Kindle. What's more, if you used MultiMarkDown it can emit LaTeX; given its footnote and endnote support, it may be a very useful tool for preparing academic papers that need a final production pass in LaTeX (a horrible format to work with by hand, in my opinion).
This isn't a formal review: it's just a comment to the effect that Scrivener works pretty much from the moment of conception to the hour before final submission of a finished manuscript. It doesn't completely replace the word processor in my workflow, but it relegates it to a markup and proofing tool rather than being a central element of the process of creating a book. And that's about as major a change as the author's job has undergone since WYSIWYG word processing came along in the late 80s (actually the late 70s if you were a researcher at Xerox PARC, but the rest of us had to wait). My suspicion is that if this sort of tool spreads, the long-term result may be better structured novels with fewer dangling plot threads and internal inconsistencies. But time will tell.
PS: Comments are still switched off due to spammers. If you want to discuss it, the Google Groups Antipope storm refuge is open for new members and I'll start a topic thread there.