Re: SpamProbe 1.2 released
Brian Burton <[email protected]>
| Newsgroups | gmane.mail.spam.spamprobe.general |
|---|---|
| Message-ID | <[email protected]> |
Anthony Campbell wrote: > On 22 Jun 2005, Jem wrote: >>>The 32 bit hash database is now "official" so I can't really change it >>>anymore. I'd like to add a 64 bit hash file option later but I've been >> >>I've been using v1.1x7 with your new Hashed-array-cached database for 3 >>months and have not seen any problems. I am using a fixed 16 MB database >>and the accuracy is just as good as before, with a much larger PBL db. >> > > So am I, except that I used a 32MB database. But I see that Brian's > advice is to use pbl in general so I'm thinking of changing. I'm a bit > worried about running out of space eventually. The primary reason I recommend PBL over hash if you have enough disk space is that with PBL there is no chance for ambiguity in the database. The hash stores 32-bit hash code not the terms themselves so you can get collisions. This is most dangerous if it happens that a message digest (used by SP to remember how it counted a particular message's terms) collides with another word and you decide to retrain that message. Is that really going to be a problem for people? So far I haven't seen any problem with this in my own database. In my recommendations in the readme I try to be conservative so that average users don't get in over their heads and yell at me. :-) The best of both worlds is probably the split database type. That uses a hash file for the message terms and a tiny PBL (or BDB) database for the message digests. I might switch to that for my own email. All the best, ++Brian ------------------------------------------------------- SF.Net email is sponsored by: Discover Easy Linux Migration Strategies from IBM. Find simple to follow Roadmaps, straightforward articles, informative Webcasts and more! Get everything you need to get up to speed, fast. http://ads.osdn.com/?ad_id=7477&alloc_id=16492&op=click