Rob Mangiafico <[EMAIL PROTECTED]> writes: > We have been running SA 2.55 for quite some time for a large user base > without Bayes support. It has worked great, but as of late, more and more > spam is getting through (as expected) and we are looking to upgrade to > 2.60. A few questions: > > - Our users probably are not going to be able to feed spam/ham into the > learning system in SA manually. Based on this, would you recommend turning > on auto bayes learning per user, or leave it off as we did in 2.55?
Its should general improves results, so if you have the resources to run Bayes, it helps. Network tests also help. > - Has anyone noticed any severe problems with bayes auto-learning with NO > manual corrections? I'll let someone running a large site with Bayes answer this one. > - Any major drawbacks to upgrading to 2.60 now, is there an imminent 2.61 > release that fixes big problems, etc...? Nothing major. A fair number bugs will be fixed in 2.61. Maybe you want to try a smaller trial of 2.60 or 2.61 (either) before upgrading everyone. (I'd do that for any large site, myself). > 2.55 with AWL and NO bayes performed great for its duration. Of > course, we are always looking to improve the system, and look forward > to any feedback on these issues. Daniel ------------------------------------------------------- This sf.net email is sponsored by:ThinkGeek Welcome to geek heaven. http://thinkgeek.com/sf _______________________________________________ Spamassassin-talk mailing list [EMAIL PROTECTED] https://lists.sourceforge.net/lists/listinfo/spamassassin-talk
