| Newsgroups |
gmane.mail.ifile.general |
| Message-ID |
<[email protected]> |
I forgot to mention one important detail: all of those timings were
for an idata database with upwards of 100 categories.
Re-running the experiments using an idata database with only two
categories (spam, non-spam), speeds things up by a factor of 3-4 on
non-core-dumping runs; in practice, that's quite usable where the
other speed was really annoyingly slow.
So daemonization would definitely render even 100-way categorization
usable speed-wise. ( Although honestly, in the long run I think I'm
going back to rules as my first line of attack, since there are too
many places where only the name of the mailing list really
differentiates message-content between groups. But I might still use
the auto-classification to handle things the rules don't...)
Jeremy