matt@srs.UUCP (Matt Goheen) (04/09/88)
This rarely happens here, but one of our users recently sent out
a mail message with LOTS of recipients. Unfortunately, this
generated multiple copies of the data file. Being a (somewhat)
conscientious network user, I was wondering WHY? I have the
'm' flag set in the "sendmail.cf" file (for "smail" -- our external
mailer). Does this option only work if the users are on the same
host? All the messages first get routed to our only mail link
(rochester). 75% of the remaining messages then go on to rutgers.
Is mail not smart enough to break up "To:" lists when mail paths
diverge? Why not (not that I volunteer mind you)? Since the above
mentioned user may be setting up a mailing list in the near future,
this seems like a good time to solve this potential problem -- or
ignore it if there is no fix for it, but at least I've done my best...
BTW, what, pray tell, does the 'm' flag REALLY do -- have multiple
addresses only for the local host? Only for messages where the
path is identical (i.e. destination host is the same)?
--
- uucp: {rutgers,ames}!rochester!srs!matt Matt Goheen -
- maybe-net: matt@srs.uucp S.R. Systems -
- "We built some good machines, but they don't work no more." -jeff@tc.fluke.COM (Jeff Stearns) (04/14/88)
In article <778@srs.UUCP> matt@srs.uucp (Matt Goheen) writes: >This rarely happens here, but one of our users recently sent out >a mail message with LOTS of recipients. Unfortunately, this >generated multiple copies of the data file. .... Matt, recall that sendmail passes each recipient address through ruleset 0 to find a {mailer, host, user} tuple. Tuples are sorted by mailer and host. You say that you're using smail, so it's likely that all your tuples resolve to mailer==smail. But be sure that all hosts resolve to the SAME CONSTANT STRING. You might expect that an address of the form "matt@srs.uucp" should resolve to {mailer=smail, host=srs.uucp, user=matt}, but this is wrong. Instead, ruleset 0 should resolve this address to {mailer=smail, host=SomeIgnoredConstantString, user=matt@srs.uucp} Now sendmail's sort phase will respect the "sumtiple recipient" flag you mention. Aside: Sendmail has changed for the worse in recent revisions; more than ever it insists upon passing TRANSPORT addresses through rulesets for HEADER addresses. This point is a bit subtle and counterintuitive. A close inspection of 4.3BSD sendmail' internal ruleset invocation will show that for most configuration files the user part should contain the complete domain address (not just the local-part). Thus the solution I propose to Matt isn't merely an optimization. -- Jeff Stearns Domain: jeff@tc.fluke.COM Voice: +1 206 356 5064 If you must: {uw-beaver,microsoft,sun}!fluke!jeff USPS: John Fluke Mfg. Co. / P.O. Box C9090 / Everett WA 98206