Last modified: 2012-09-22 17:40:48 UTC
Translation memory is a much requested feature of the Translate extension that is used in translatewiki.net but not in Wikimedia projects.
The proposed shared database layout  was found to be unacceptable because it uses MyISAM, which is no longer allowed on the cluster.
I am now researching whether we can implement the fulltext filtering via Lucene. I've looked at few classes at MWSearch for example, and I'm not enthusiastic of what I've seen.
Should also check whether the performance is acceptable without fulltext filtering.
I've been talking with Oren to get some help in solving this.
Totally forgot this bug. Finally I can close it :)