I have been implementing a system-wide filter. You can do a reasonable
amount with the filter language (see F00* in the config.samples file
available for download from the main site) but there are many limitations.
Obviously Philip didn't set out to build another awk, perl or procmail.
If you try to do anything at all structured the read-only nature of the
filter processing language becomes a bit painful. There are only 10
variables and these are counters, and the ability to set values for
subject lines etc is confined to parameters in a few commands like 'mail'
and 'log'. It would be nice to be able to store the path for all database
files in a single global variable for example, and to be able to set
$header_* variables at arbitrary points. Embedding Perl doesn't seem to
address the problem although it makes processing much more flexible within
a single expansion.
My question is this: if I deliver all mail first to a procmail rule as per
the example in spec.txt what do I lose through not doing this in a system
filter? I would have to set an X-loop: header and forward to myself after
doing all the processing, and procmail doesn't (I think) have any way of
tying into the concept of significant deliveries like the filter language
does.
Thanks,
Dan
dan@???
--
*** Exim information can be found at
http://www.exim.org/ ***
