On Thu, Jul 03, 2003 at 10:46:32PM +0100, Giuliano Gavazzi wrote: > >I would,, respectfully, disagree :-)
> >I'm not talking abuot caching across multiple children, but caching
> >during the length of one connection, while you're trying to determine
> >the route of the message.
> >
> >This is what I believe happens at present with local_domains specified
> >in the config file. I'm only suggesting we do it for database queries
> >also. It'll make it more consistant across the different types of
> >lookups, and reduce the number of database queries required.
>
> only one problem, how does it know that you did not update the table?
> Clearly in your example that would not make much sense, but how does
> exim know? Perhaps make it an option? (or maybe that is already an
> option?) Did I write a single sentence without a question mark? No.
It shouldn't need to. Infact, if anything, it *SHOULD* cache data for
the length of one connection, because updating that data could in
theory, break the delivery.
It only has to last a few seconds, while the current mail is delivered.
Then the next processes does another lookup to get the data for itself.
Why on earth would you WANT Exim to get one piece of information at the
start of a delivery, and a different piece of information half way
through? I can't think of a signle situation where it would :-)
However, if you can convince me that the current behaviour with database
lookups (to be non-caching) is correct, that I would state that we
should remove *all* such caching from Exim, for the sake of consistancy.
