I just deleted over a million worth of backlinks from our database, namely all becklinks from 2002. Our policy of keeping per-story backlinks forever just did not work out - the accesslog table is (or was) one of our biggest headaches. It causes a lot of the slow queries for MySQL, and when I checked I found it was corrupt again today. I think in the future we'll have to adopt a different policy of keeping story backlinks for a month or two, max.

During the time MySQL took to delete the records, access to individual stories timed out because the the table was locked. Sorry for that outage and the ones before which were caused by other db maintainance work.

well, i always wondered that there is something like an automated cleanup mechanism built into antville, which automatically deletes old accesslog entries, but that it is not used here. lets keep the fingers crossed, that this was it! another performance issue might be the storiesByID collection, which is only used for the bloggerapi, and therefore could be easily substituted with a direct database-query. (having that said, i will try to rewrite this part and submit it to the CVS if that's fine with you guys)

I simply don't understand what you like to archive with the backlinks-tag, it sounds to me like a sysiphos-job because:

• it works against every cache-strategy to modifiy a story when somebody accesses it with a valid referer.

• it accumulates lots of data, that is quite useless within a story.

Don't get me wrong, I am not saying to skip it, but maybe its a wise decision to implement it like the /referrer/ on a per story basis like project.antville.org?