A few days ago we changed our website infrastructure from a Confluence export that gets rsync'ed over several hosts until it reaches the ASF's main web servers to a SvnPubSub based approach. Now once an hour a buildbot build exports the site from Confluence and checks the changes into SVN where they get picked up by the public facing web servers almost instantly. This allows us to keep our old website sources in Confluence but also allows us to check in static resources like JavaDocs, XML schema files, DTDs, etc. directly into SVN and see the changes almost instantly. In addition, the new buildbot-run export cleans up almost every reference to Confluence, thus speeding up site rendering.
We also took the opportunity to clean up some of our legacy stuff. We took offline all documentation to the many previous versions of Tapestry and left only the static documentation (i.e. without the JavaDocs and reports) for Tapestry 3, Tapestry 4.1, and the latest version of Tapestry 5 online. If you require the documentation for another version of Tapestry, you will have to download the archived sources from http://archive.apache.org/dist/tapestry/ and build them for yourself. The JavaDocs and component reference for the current version of Tapestry 5 are linked from Documentation.