Re: [EXIM] System Filter Efficiency

Startseite
Nachricht löschen
Nachricht beantworten
Autor: Nigel Metheringham
Datum:  
To: Tony Earnshaw
CC: Exim
Betreff: Re: [EXIM] System Filter Efficiency

tony@??? said:
} Anyway, what I wanted to say, was that the file has now around 1.300
} sites, that Exim has to read every time he receives mail. (He also has
} to do gethostbyname()s and RBL lookups on all of the incoming mail
} servers, _and_ he runs from inetd (TCP Wrappers connected). All the
} same, it seems to me that he runs as fast as lightning on our SCO
} 5.0.4 box (1 Pentium 166) - far faster than MMDF which I (and many
} customers) had before.

So make it a keyed lookup - you wouldn't believe how fast a cdb lookup is.
Certainly lots of this stuff can easily be done in the many lookups that
are available, rather than dropping to the filter - which is always going
to be a lot slower than the native lookups.

The real problem is that a load of people are now trying to do some really
complex stuff - weighted combinations of spam keys, and "if connection is
from x then filter with y else filter with z", which can just (or can't)
be done in exim's filter and certainly cannot be handled by the config
variables.

The whole thing again starts to feel as though some form of internal macro
language is needed, with reasonable capability. However re-implementing
yet another language is rarely a good thing to do. Integrating another
scriptable form would have its big advantages [wave to Malcolm Beattie at
Oxford in the hope that the exim-with-embedded-perl magically appears].

Maybe I'm too close to the problem and a deep breath and good step back is
needed. People are starting to do some very interesting stuff which
really needs a different paradigm (buzzword alert).

Also there have been some discussions around on performance - how to
squeeze serious performance out of an MTA. Now exim ain't at all bad.
However its not in the same league as (say) apache for designed for
performance. A load of the apache optomisations could be applied to an
MTA - pre forking or even threading etc. However exim will have problems
here as long as it fork/exec (followed by startup and config read) to
regain priv during delivery.

However at present I could be persuaded that exim has enough performance -
its much better than many other implementations - and if the performance
impacted the security model, that could be hard to swallow. Its if you
want to do full message analysis that things get hard at present.

    Nigel.
-- 
[ Nigel.Metheringham@???   -  Systems Software Engineer ]
[ Tel : +44 113 207 6112                   Fax : +44 113 234 6065 ]
[      Real life is but a pale imitation of a Dilbert strip       ]




--
*** Exim information can be found at http://www.exim.org/ ***