Brett_Tabke - 1:21 pm on Apr 19, 2010 (gmt 0)
StoutFIles - core db is modified Sphinx. There is another meg of our Perl code for a spider and ranking algo. Sphinx is awesome in that it doesn't require an sql db. We just feed it raw xml. However, there is a major ton of tweaking to get the level of relevance that we want.
> format the output more like the SEs.
Agreed - however, there are significant differences in the way we have built this SE vs the way big G/Y/Bing work. We are indexing down to the message level and yet, we need to return SERPS at the page/thread level. That leaves a wide array of interpretations on how the display should look. We also want to be able to toggle between two modes (thread/page mode vs message mode). Lastly, we are going to throw in a 3rd option and just allow people to open the message right on the SERP via ajax.
So, we are going to monkey with it a bit more this week.