amix: pretty nifty what they did, and apparently it scales for them

i really doubt it scales that well on many updates or many objects

they have wikipedia indexed and it has an acceptable performance

i'm not quite sure how it would *scale* worse than xapian. it just puts more information into the btree

mitsuhiko: Xapian scales pretty badly based on my experience (we had Plurk user search done in Xapian first)

that said, i think Xapian is "ok" solution if one does not do that many updates [i.e. for fairly static database]

all search systems have their advantages and disadvantages. xapian does pretty well if you can live with slightly delayed search

eg, no real-time indexing

i agree, i just don't think delays are that good in an object database

the object database is intended to be used for search relevant things only

basically the idea is that you can use xapian only without having to fire a single sql request for search result filtering

ok, fair enough. you could try to use it in Solace, i think it might be a good solution as the data size won't be that big