Eric Astor wrote:
On this note, would it be possible to get a public pointer to the full skins, modifications, and so forth used in creating the latest static dumps (as at http://static.wikipedia.org)? Several projects I know of (including the OEPC project I'm working on) are looking for ways to create good static exports, and the November dumps come out looking excellent.
All the scripts used are public, they're at
http://static.wikipedia.org/scripts/
This is my actual working directory for constructing the November dumps, I decided that I didn't need to shut it away. The main entry point is do4thread. All patches and modifications were committed to CVS.
The skin used is at skins/disabled/HTMLDump.php, you need to move it to skins/HTMLDump.php for it to work.
In a second post, Eric Astor wrote:
Not quite... I've done a dump, and the format comes out differently - not really appropriate for a static dump, since among other things, the Search button is present but fails.
The JavaScript referenced by this skin is in skins/htmldump/, including lookup.js for the "go" button. It's not really a search button, since I haven't gotten around to writing a fully-fledged search engine in JavaScript just yet. It doesn't handle misses particularly gracefully.
Brion Vibber wrote:
I would be happy to assign a server for conversion duties. Originally srv31 (which hosts static.wikimedia.org) mostly did this; though it's been kind of taken over for additional things they can be moved.
srv31 would probably best, if it's got the hard drive space. dumpHTML is CPU-intensive, so it would be best if we could move other CPU-intensive tasks off it.
Actually what would really be nice is a server on VLAN 1. I wanted that last year but didn't get it. Instead we have srv31 serving webpages via squid, and downloads via NFS and benet. It would probably be better to do both directly.
-- Tim Starling