I'm trying to set up experimental site for search.debian.org by hyperestraier 0.3.13. This indexes only html files in http://www.debian.org/ . It is only 30000 files or so, so index file size is around 254M (note that www.debian.org mirror size is 1.5G).

Before this, I also tried to set up http://fabre.debian.net, which is search engine for http://bugs.debian.org/ . There are huge number of messages in BTS, so that it is hard to serve on this PC (PIII 600MHz 512MB). I also experienced that it took so much time to create index database. It would be better to customize and tune up more. For example, we don't need control message, except after "thanks" keywords, in search index. It only has bug reports by July 15th, so bug reports in this month are not indexed.