from the no-more-big-database dept.
In the last months, the bayesian database grew up to 9 gigabytes. This amount of data is a problem to handle and the benefit of a personal collection of bayesian tokens shrinked during the last months. So we decided to switch to a global collection for all users.
Users with a small amount of email will benefit from the lager database; users with a large amount of atypical mails may get more spams. For those users, we recommend the usage of the built-in spamfilter in most modern mailprograms like Thunderbird or Apple's Mail.
The switch was motivated by the time to recover the whole database (around an hour), restore from the backup (arond 5-10 days) or convert to InnoDB for faster access (more then 20 days). On the other hand, the benefit is no longer the same as one year ago - most spam is caught by the simple rules and sophisticated spam wont be recognized by the bayesian filter.
< | >