Re: consistant failures on a new kind of spam
Brian Burton <[email protected]> Wed, 15 Nov 2006 12:00:25 -0500
| Newsgroups | gmane.mail.spam.spamprobe.general |
|---|---|
| Message-ID | <[email protected]> |
-----BEGIN PGP SIGNED MESSAGE----- Hash: SHA256 Jem wrote: > Brian I thought that the built-in hash database (Hashed-array-cached) had a > fixed size, presumably because a hash collision just overwrites a hash > entry. So I thought whether you clean it or not, a collision will lead to a > term being overwritten. Or are you saying it grows to a list at collision? Oh, and collision can mean one to two things. In my email I was referring to two different hash codes mod'd to the same index into the array. I think you were referring to the second (and nastier) kind of collision which is when two different terms hash to the same hash code. In the latter case you are right that a term can wind up being overwritten due to a hash collision. I think my testing of the hash algorithm showed a very small number of this type of collision in my own email. The former case (two codes mod to same index) is harmless except that it impacts performance as I described in my other email. And you are correct that neither case will cause the file to grow in size. The hash files are fixed size and can become too full to hold new terms. Thanks, ++Brian -----BEGIN PGP SIGNATURE----- Version: PGP Desktop 9.5.1 (Build 1557) Charset: ISO-8859-1 wsBVAwUBRVtHrDxRyEoJfXIFAQizXggAsOH6Ezo93w2yWg1kkuwqddoemTELYZeO 8lYSl4tQw3a08Zex6UX6VCIiwiplI2pVWnMpsvZV5DXIolPxGd1pPQkuWPyz29Ir cWXC8JepWyakhc63XpY1qDQXqEsu1uHTtB2RrnW4bGAfWoFe6rBy/rSzvXW8dfIZ XVRgdFMe3dG4pbZeaOEm5hXSa/+PP77EQ2XjC+EEEUfAyZLnNAubU5HgDpgsSw6e dYOLpLPqhCRFO2O7bD4l4Oj3csC4xB4PnRzyMiQw+XS1U2Ixpf/as1jRuzMzbCpm oDXDappe1fQXsW6r2FDFlvCI5iKd9zj8EjFxMZOPPAt7x6jIbh/L3w== =V2AT -----END PGP SIGNATURE----- ------------------------------------------------------------------------- Take Surveys. Earn Cash. Influence the Future of IT Join SourceForge.net's Techsay panel and you'll get the chance to share your opinions on IT & business topics through brief surveys - and earn cash http://www.techsay.com/default.php?page=join.php&p=sourceforge&CID=DEVDEV