I'm happy to say I'm now a Forrest committer! All it took was a few rants, some semi-intelligent ml discussion, and one monster patch which nobody felt like applying :) Forrest has masses of potential, if only it can focus on being a generic doc system instead of an xml.apache.org revamp.

Since then I've been working on making Forrest as braindead simple to use as possible. I'm pretty happy with the progress so far. Once the Forrest binary is installed, you can create and render a template site by typing 'forrest seed site'. To generate a webapp all ready for deployment, type 'forrest webapp'.

 I wrote up a Forrest "getting started" guide at http://xml.apache.org/forrest/your-project.html. If you're looking for an XML->{HTML,PDF} doc tool, have a look.

I've also written a [RT] Linking revisited: A general linking system, with some ideas about how Forrest xdocs could link to each other. Currently, each page links to the HTML rendition of other pages, eg <link href="foo.html">. That is conceptually broken; why should a mere link assume what the sitemap is going to render foo.xml as? What happens if foo.xml moves to a different directory? Anyway, the RT proposes a system whereby we could write <link href="site:site/foo"> instead, meaning the 'site/foo' node in an abstract tree of nodes (a node is a link-to-able bit of site content). Implemented with Cocoon Sources, fancy stylesheets and a bit of hand-waving. Thoughts on it are welcome.

Robert Koberg has implemented some very nifty stuff for his LiveStoryBoard CMS. The site looks plain, but poke around at it's internals and you'll see some pretty amazing use of XSLT. In particular, LSB makes heavy use of a central site.xml file, similar to what my RT proposes for Forrest. Robert kindly sent me a copy of LSB 2.0 (the site there is 1.0) which I'm looking forward to having a poke at.