Re: [Exim] ANNOUNCE: exiscan-acl-4.24-13

Página Principal
Apagar esta mensagem
Responder a esta mensagem
Autor: Tom Kistner
Data:  
Para: Michael Haardt
CC: exim-users, eximusers
Assunto: Re: [Exim] ANNOUNCE: exiscan-acl-4.24-13
Michael Haardt wrote:

> I am not absolutely sure, just pretty much so, in particular if I remove
> the "true" option, which nobody proved so far not to be a feature Exim
> can perform without additional code already.


Yes, but it would make the ACLs look clumsy in most cases. With the
'true' option, you can say:

deny  message = Spam score above 8 point cutoff, rejected.
       spam = joe:true
       condition = ${if >{$spam_score_int}{80}{1}{0}}


Without it you'll end up with this:

warn spam = joe
      condition = 0
deny message = Spam score above 8 point cutoff, rejected.
      condition = ${if >{$spam_score_int}{80}{1}{0}}


> Honestly, I don't care what Debian does. I enjoy free software,
> because I am free to use and develop it. So far, nobody expressed
> feeling miserable about me hacking the source and distributing patches,
> but there is always a first time.


The installed base of exiscan-exims is now VERY large. The patch is in
packages of FBSD, Debian, Gentoo, SuSE and there are numerous RPMs
available. There is absolutely no way I can implement backward
incompatible changes.

> I do not force anybody using my code. If you don't like it, then rewrite
> it, or have people download the source, apply the patch and wonder why
> it is not in the package, as a bunch does right now, according to the
> downloads of it. Either way, not my problem.


That is correct, however the general public would benefit more from
backwards-compatible changes.

> How about you providing a patch that introduces attribute-value pairs,
> so we can specify a timeout in future, too, in order to rather delivering
> mail unscanned than deferring it if the scanner is down, for those who
> prefer that? It would break compatibility at least only once, and for a
> better reason than my code.


To return to a technical discussion, how about this:

warn message = blah
      spam = joe:true:spamd=127.0.0.1 783


... where both the 'true' and 'spamd' options are of course optional.

This is backwards compatible, and hey, you can have 'timeout=5s' and
'error=ok' too if you like. The username will have to stay where it is
of course.

/tom