Ask Bjørn Hansen said the following on 02/24/2006 08:39 AM:
> 
> On Feb 23, 2006, at 1:27 PM, Robin Bowes wrote:
> 
> [using plugin inheritance]
> 
>>> The only thing I haven't worked out is how to handle any optional
>>> commandline arguments to the original queue plugin (which may just  be a
>>> matter of calling the original plugins init() sub with the 
>>> appropriate @_).
> 
> 
> Yeah, that should work.   We really need a better configuration  system
> for the plugins though.
> 
>> That's certainly another approach.
>>
>> However, it introduces unnecessary coupling between the logging and
>> queueing processes. I'd rather keep them as separate as possible.
> 
> 
> Isn't this approach exactly a way to keep them separate?   The other 
> suggestions so far have involved changing the queue plugin(s) which 
> seems more intrusive...

I think the time is right for a global change here as John has proposed,
i.e. foo_pre, foo, and foo_post hooks.

> Thanks for working on this by the way.  I lost track of what your  goal
> was, but having more and better way to analyze how much junk the  smtpd
> rejects (and how and why) will be useful.

My goal is (was?) relatively simple: to produce two log files in
addition to the "normal" log:

accepted - listing details of all accepted msgs
rejected - listing details of all rejected msgs/connection attempts

> Robert mentioned a week or two ago that the perl.org (etc) MX only  ends
> up accepting a mail from 20% of all connections and of the  accepted
> mail most ends up being tossed in a huge file never to be  looked at
> again.  (So something like 95%+ of the connections end up  being a waste).

Well, I'd like to be able to easily produce statistics like that.

My gut feel, based on my experience running qpsmtpd for a medium-sized
ISP (~350 domains) is that those figures are under-estimates.

As an aside, isn't great to be able to easily modify the code like this?
Can you imagine trying to patch qmail-smtpd to do this? Kudos.

R.

Reply via email to