Provide a translation memory, both for internal an external use
Submitted by Gil Forcada
Translation memory: http://en.wikipedia.org/wiki/Translation_memory
A member of the Catalan team was updating our own home made based translation memory and run it over some of our translations to found out that around 2k (yes 2000) strings could be translated by the translation memory itself (so 100% coincidence).
I think that it would be really great to have a translation memory generated by Damned-Lies itself and that it was used when updating po files to also pre-translate (marked as fuzzy) the 100% coincidence strings.
This would mean, that for example, translating, say rhythmbox and pushing the translation to git and then downloading banshee to translate it will have quite a few strings already pre-translated that will only have to be reviewed, instead of fully translated (lookup the vocabulary, consistency..., it's a lot of wasted time that could be spend translating other strings).
This translation memory could be updated daily/weekly or on-the-fly (will have to look at performance).
The external use will be to either allow downloading the translation memory, or provide a simple API that gtranslator/virtaal/kbabel(or whichever name has now) could query to get results from (much like open-tran.eu but only for GNOME translations).