On Mon, 2004-11-22 at 02:49, A. Lanza wrote: > I read an article explaining how to integrate SpamAssassin with > Evolution 1.4 and decided to try it. I'm running SpamAssassin as a > daemon (spamd) and filtering mail using a rule that pipes the messages > to SpamAssassin client (spamc -c) and checks for the return code to > clasify messages as spam or not. > > I've noticed that fetching mail is much more slowly this way than > before, but it's not because of processing time taken by SpamAssassin, > processor usage stays low. Anybody have an idea of what could be > happening? I would recommend that if you are processing every message you get including mailing list messages through spamassassin that you change the filtering processes to skip mailing lists. I have seen very few spam type messages come across mailing lists. Setup your filters to identify mailing lists and stop processing. Then have the filter that calls spamassassin. It will only be invoked on messages that more likely to be spam. I think you will find your mail fetch process will run much faster. Don't forget to put the stop processing rule in the mailing list filters. As to what the problem is. Even if you are using spamd (which is more efficient that calling spamassassin directly every time) evolution still has to load the script called by the filter for each message and invoke spamc (another load) on each message. In a typical MTA setup you would have sendmail (or the MTA of your choice) running some kind of milter that would be already loaded all the time. When a lot of messages would hit it it would sequence through the messages without having to load the program each time. That is probably what is causing your speed problem. -- Scot L. Harris webid@xxxxxxxxxx In every non-trivial program there is at least one bug.
