Re: [Exim] message_body efficiency hack

Top Page
Delete this message
Reply to this message
Author: Nigel Metheringham
Date:  
To: Dr Yann Golanski, Exim
Subject: Re: [Exim] message_body efficiency hack
yann@??? said:
> It's realy hard to say, but remember that Exim is very IO limited and
> as such bigger header files are bad. However, I think it is worth
> investigating.


IO limitations are more to numbers of files rather than bulk of
files... ie its hitting an extra file which is really painful, or
blowing your file cache stuff so you have to go back to the spinning
rust. However point taken.

> A potential problem is that the header lines might be passed along
> with the messages, which you obviouslky don't want ;>


huh?
This isn't a real problem - you are just adding a couple of variables
which are set at receive time and stashed in the headers file - like
lots of similar variables.

Sheldon Hearn <sheldonh@???>:-
> Are you proposing to impliment this as optional behaviour or will
> everyone have to pay the performance penalty regardless of whether
> they need this new functionality?


If you set message_body_visible to zero then it would not save this
data. Its worth noting that message_body_visible would take effect at
the time of message reception - if yesterday (for some strange reason)
you added a filter, upped the message_body_visible from zero, then
messages in the queue before the change would not have the body
available.

I would be unhappy having 2 separate chunks of code to do the same job
in different ways :-)

    Nigel.
-- 
[ - Opinions expressed are personal and may not be shared by VData - ]
[ Nigel Metheringham                  Nigel.Metheringham@??? ]
[ Phone: +44 1423 850000                         Fax +44 1423 858866 ]