Hallo list,
I just took a look at the code.
AFAIU the complete revisions are stored in a compressed form in the text table. Why did you choose this way instead of just saving diffs between revisions like versioning systems like cvs and svn use to do.
Wouldn't that be more memory (and money;-) economical?
Bravo for your work, Cyril
ps: Sorry if that question might have been posted before, but I found no search feature in the archives.
___________________________________________________________ Telefonate ohne weitere Kosten vom PC zum PC: http://messenger.yahoo.de
On Thu, May 11, 2006 at 10:38:11AM +0200, lyric680-web@yahoo.de wrote:
Hallo list,
I just took a look at the code.
AFAIU the complete revisions are stored in a compressed form in the text table. Why did you choose this way instead of just saving diffs between revisions like versioning systems like cvs and svn use to do.
Wouldn't that be more memory (and money;-) economical?
Actually, the revisions are only saved like that for a certain time, until compressOld comes along (every few months). CompressOld takes several revisions, concatenates them and zips them. The size of this operation is not much different from a full version and a set of diffs.
But it's faster. For an article like George W. Bush with many thousands of edits, it would take ages to compute older revisions. You'd have to apply thousands of 'patches' to the current article to get the old one.
Regarding economics: The texts of the revisions aren't stored on our expensive high-end database servers, but on a farm of cheap servers that are also used as apaches.
Regards,
jens
wikitech-l@lists.wikimedia.org