a brief tutorial on sendmail rules

[Greg Earle]

> While I am running off at the keyboard here, here are some of the
> other things I would also like to see in sendmail.  Remember that
> this is supposed to be an address rewriting programming language.
> (What else can you call it?)
> 
> 	- Long variable names.  32 characters is an absolute minimum;
> 	  flexnames preferred.  Small name spaces are bad.  Remember
> 	  the *big* 8K (core of course) computers you worked on?
> 
> 	- Real control structures, not just `do this once per
> 	  match' vs. `do this every match'.
> 
> 	- Table lookups with return values, tables usually
> 	  being files of some kind, hopefully fast (e.g., hashed).

To which I'll add my two cents worth:

	- FIX THE PROBLEM WITH LONG DESTINATION ADDRESSES!
	  *Three* times I have seen mail come in from a site 1 away
	  from us, which was intended to be forwarded by my machine,
	  where (because of the incoming news path) the destination
	  address was like 19 sites long.  Sendmail (ala 'sort'),
	  "silently truncated" the long dest. addr to 1!2!...!14!15
	  so it looked like the destination was user (machine 15)
	  at site 14!!  In all three cases, the destination address
	  got cut off at 105 characters long.

	- either extend the 'F' class operator, so that it can return
	  a 2-tuple, with key and associated string, or use a new
	  letter that can perform this feature.  I am pretty sure I
	  saw this in the Maryland sendmail changes; it should be
	  standard.  There's no reason at this late stage in the game
	  why everybody should have to have a 'uumail' program written
	  especially to sit in between sendmail and uux just so you
	  can use the pathalias data.  Sendmail should do this itself,
	  via this changed/new operator.

	- There seems to be an oversight in Ruleset 0, whereby if you
	  don't put in your own rule to strip off the name of the relay
	  host when sending to the Arpanet from a uucp site, then later
	  on if the receiver tries to reply, the address will be such
	  that it thinks it has a uucp link to the Arpa site that first
	  got your outgoing mail (instead of a TCP link).  The reply
	  gets munged ... Our rule for this (Thanks to David Robinson
	  at Caltech) is :

# Remove name of relay host if sending to arpanet
R$R!$*<@$*$-.arpa>$*    $#$M    $@$R $:$1@$2$3.arpa$4   rhost!user at any.arpa

---------------------

	Greg Earle
	sdcrdcf!smeagol!earle (UUCP)
	ia-sun2!smeagol!earle at csvax.caltech.edu (ARPA)