Last modified: 2013-06-18 16:20:38 UTC
A special version of upload.tar which only includes the most recent version of
files would be nice. Perheps only including files that are linked from the
current version of the articles.
Regular upload dumps are now provided, marking this as FIXED.
Reopening: This feature-request is not about a recent dump but about a dump
containing only the most recent versions of files.
While actual .tar files are probably not feasible at our current level (~3TB for Commons files current versions only), getting some offsite image mirrors and redistribution is on the table. Tomasz, assigning this one to you since you'll be coordinating the data dump stuff.
Releasing this bug so that anyone who has time can take it on.
These are now semi-available (I'm running them on an ad hoc basis, they are generated on a mirror site rather than one of our servers, we're still working out hardware issues with them, etc etc.) If you're willing to deal with directories moving around and possible inaccessibility, you can get these before the official announcement, from http://ftpmirror.your.org/pub/wikimedia/imagedumps/ in the tarballs/full directory and the tarballs/incrs directory. These are indeed current version only, per project *except* for commons.
If you want commons images, you should get them via rsync from rsync://ftpmirror.your.org/wikimedia-images/ and please see http://meta.wikimedia.org/wiki/Mirroring_Wikimedia_project_XML_dumps#Current_Mirrors for more information about what data is mirrored where.
Anyone on this bug that's not on the xmldatadumps-l list had better get on it, since that's generally where updates about this sort of thing will be sent.
Hmm I guess since the official announcement went out we can call this done, or close enough to done at any rate. (Everyone on the xmldatadumps-l list yet??)