i gave up on mnogosearch after experimenting with it. i tested it on
the Alpha using NetBSD-current. after the mnogo team fixed some bugs so
it would run in cache mode on NetBSD w/o as many problems, there still
were problems if you index over a million urls. splitter started to
core dump, it used to coredump at around 20K urls before the fixes.
this was back at version 3.1.12. i still see problems reported. i don't
believe the problem i reported back then when i got over 1 million urls
were fixed. never did see a reply to the post with my data file posted
on my website.
for me the only real usable mode was cache mode. it was fast. the other
modes just won't do for search with a large number of URLS. in addition
to the core dump problems above, i found the search results weren't
very accurate. phrase searching didn't really work as "phrase search"
but more like records with both words in them. relevant results weren't
ranked very high. in cache mode, i can't have 256 char width for
categories, instead i'm limited to 32bit integers. sql supports large
width categories, but that's to slow to be useable. i use google
regularly and i'm just so surprised by their results each time. that's
amazing i think.
the mnogosearch team is pretty incredible and extremely responsive. i
think alex, glux, and max are really attentive to user's questions.
admittedly, there is only so much they can do with limited time. i
think they have a potentially great product on their hands, but it does
still need lots of work before it matures.
__________________________________________________
Do You Yahoo!?
Get personalized email addresses from Yahoo! Mail
http://personal.mail.yahoo.com/
___________________________________________
If you want to unsubscribe send "unsubscribe general"
to [EMAIL PROTECTED]
