Re: [cobalt-developers] Sendmail weirdness excessive recursion

[Jeff Lasman <jblists@xxxxxxxxxxxxx>]

Taco Scargo wrote:

> Clear DNS issue: All cobalt RaQs foward mail to the full hostname, so if you
> have a site www.domain.com with e-mail server alias 'domain.com', all mail
> that is received for domain.com will be forwarded to www.domain.com where
> all the users reside.
> When you have a look at the dns file, you see there are no MX entries for
> www

Thanks for finally explaining this simply enough so even I can
understand it <smile>.

I should have realized it of course; I've known of the RaQ using "www"
for email for years, and I've even explained it to lots of people
myself.  But, when I saw a CNAME for the "www", I never connected it
with the forward of the email in my possibly overtired brain <wry grin>.

I'll explain it to the client today.

> So sendmail will then try to use an A record. As querying for an A record
> does not work (sendmail's fault for not handling cnames correctly)

Yes, sendmail's not handling CNAMEs properly is the basic reason, but
still, the Sun Cobalt method of forwarding to the www is the only
circumstance under which it'll fail, so I really still see it as a Sun
Cobalt issue, though I agree it must be fixed in DNS.

There's no intuitive way of anyone knowing you can't use a CNAME for the
www record on a RaQ, and in fact I've never seen any other platform in
which it would be a problem; there's no other platform I've ever seen
that forwards mail from the domain name to an arbitrary hostname without
explaining it to the end user in terms s/he can understand.

> sendmail
> miserably fails. So either get rid of the cnames or add a line like www
> IN    MX 10 @

I shall explain to the client he cannot use CNAMEs for hostnames which
he sets up on his RaQ for the purpose of receiving domain email.

Again, thanks!

> The @ by the way represents the 'origin' in this case 'domain.com'

Yes, that I know; I even wrote it in the post I made (the one your
answer referred to).  I've NEVER seen it on the right side, not in any
book, not in any example; though there's nothing in the relevant RFC
that disallows it I wasn't sure it was implemented properly.

Jeff
-- 
Jeff Lasman <jblists@xxxxxxxxxxxxx>
Linux and Cobalt/Sun/RaQ Consulting
nobaloney.net
P. O. Box 52672, Riverside, CA  92517
voice: (909) 778-9980  *  fax: (702) 548-9484