http://bugzilla.spamassassin.org/show_bug.cgi?id=3064
------- Additional Comments From [EMAIL PROTECTED] 2004-03-01 10:19 -------
Subject: Re: sa-learn should ignore "use_bayes 0" setting
> Well, the other problem with this, of course, is that this is great
> for us, but what about the people who use our modules directly? I, for
> instance, have code that calls M::SA->learn() directly. Should that work
> if the config says "use_bayes 0"? IMO, it's up to the modules to check,
> not my calling code.
See, I knew I was missing something, and I do the same thing, so no
excuse for missing it I guess. You're right, the callers should not
have to perform the check themselves.
> I think we'd also want to alert people if using sa-learn w/ "use_bayes 0"
> so that they know the learning won't be used by default, or for users
> who think bayes is enabled, but the admin disabled it ...
>
> Should "spamassassin -r" and "-k" still teach the Bayes system in
> this case, and how is it related to bayes_learn_during_report?
>
> etc ... There's a lot of touchy issues here IMO.
My overall position is that all of this is unnecessary, but I think
that might be coming from someone that finds Bayes invaluable and will
never have it turned off. I can see where you'd want to have it
turned off but still "usable," for some definition of usable, via
sa-learn, so I'm willing to talk about options.
How about something that gets set in sa-learn so we still perform the
check but it's like:
if (!use_bayes) {
if (special_sa-learn_var) {
warn "You're using sa-learn but bayes is currently turned off.";
}
else {
return;
}
The special_sa-learn_var could also possibly be another config
variable (use_sa-learn) or something like that.
Michael
------- You are receiving this mail because: -------
You are the assignee for the bug, or are watching the assignee.
