> > An MX record may not point to a C record even though it might
> > seem to make sense to do so, simply because DNS was designed
> > specifically prohibiting that.
>
> It doesn't make that much sense:
> 1. One additional lookup.
> 2. Mail sent to CNAME should be "canonicized" - ie. the CNAMEs
> in the addresses should be resolved. If MX points to CNAME, you
> don't know if it should be canonicized or not.
>
> Plus, of course, RFC says against that.
Yes, what I mean by "makes sense" is that it isn't apparent
reading a DNS record the fundamental difference between the
behavior of a macro and a pointer. An MX record points somewhere
whereas a C record rewrites.
So if you mentally translate an MX record of
family.cybergood.net. IN MX mail.bannerclub.com.
as "mail sent to [EMAIL PROTECTED]
is handled by the computer that is represented
by mail.bannerclub.com"
makes sense (both family.cybergood.net
and bannerclub.com are A records)
If family.cybergood.net or mail.bannerclub.com were
a C records (not in itself unreasonable for web sites)
the read of the MX record would still "make sense"
but be wrong.
> -----Original Message-----
> From: Petr Novotny [mailto:[EMAIL PROTECTED]]
> Sent: Thursday, July 22, 1999 3:58 AM
> To: Alex Miller
> Cc: [EMAIL PROTECTED]
> Subject: RE: Can qmail work withuot MX RR?
>
>
> -----BEGIN PGP SIGNED MESSAGE-----
> Hash: SHA1
>
> > An MX record is only necessary when an A record points to a machine
> > that has no mail handler, and the MX record then points to an A
> record of
> > a machine that does have a mail handler.
>
> That's the other way around (I mean the reasoning): Empty MX list
> is considered a special case when A record should be tried. You
> should try to have a MX (even pointing at yourself, like
> mail IN A 11.111.111.1111
> mail IN MX 5 mail
> ) because it saves one DNS lookup.
>
> > An MX record may not point to a C record even though it might
> > seem to make sense to do so, simply because DNS was designed
> > specifically prohibiting that.
>
> It doesn't make that much sense:
> 1. One additional lookup.
> 2. Mail sent to CNAME should be "canonicized" - ie. the CNAMEs
> in the addresses should be resolved. If MX points to CNAME, you
> don't know if it should be canonicized or not.
>
> Plus, of course, RFC says against that.
>
> > Sending mail to a C record might seem to work, deceptively,
> > because you may have included the A record in localhosts.
>
> ???
>
> You can send mail directly to a host which is only a CNAME (and
> the address gets rewritten to what CNAME really points at). In fact,
> if you have CNAME and MX for the same host, the MX is ignored
> (and probably considered an error).
>
> > For example,
> > if you have an A record stooges.com on a machine with QMail
> > and stooge.com is in the /var/control/localhosts file
> > then any mail sent to [EMAIL PROTECTED], moe.stooges.com,
> > [EMAIL PROTECTED] will deliver if larry, moe, and curly
> > are users.
>
> If you meant [EMAIL PROTECTED], yes.
>
> > If there is a C record three.stooges.com that aliases
> > stooges.com (that's what a C record is) then
> > [EMAIL PROTECTED], [EMAIL PROTECTED],
> > [EMAIL PROTECTED] will also deliver because
> > like a c programming macro three.stooges.com
> > "becomes" stooges.com
>
> Yes - I like that explanation :-)
>
> > The deceptive part of that is this.
> >
> > If you want to route mail of [EMAIL PROTECTED]
> > to go to curly@localhost you must creat an alias
> > for shemp to route to curly and that will work for
> > all localhosts.
> >
> > So there will be an effect that [EMAIL PROTECTED]
> > will also go to the user curly.
> >
> > For explicit control of each domain, separately, you
> > want to use virtual domains, not localhost.
>
> Sure - all in "locals" are aliases for local host; ie. after you decided
> the address in in aliases, only local part is used to route the mail. If
> you want to use different handling for different domains, you want
> virtual domains.
>
> Caveat emptor: If three.stooges.com is a CNAME for stooges.com,
> you are not too likely to receive a mail for
> [EMAIL PROTECTED], since the CNAMEs should get
> canonicized (rewritten), unless the other side is broken. If you want
> virtual domain, you want MX pointing at you, not CNAME.
>
> > Each virtualhost should have it's own A record.
> > stooges.com has an A record
> > three.stooges.com has an A record
> > littlerascals.com has an A record
>
> MX is enough. Although nothing prohibits it, you might (repeat:
> might) have some problems with broken, brain-dead machines on
> the other side that do a forward lookup, a reverse lookup, and then
> seem surprised that they didn't get the original name. But hey,
> yes, their behaviour would be broken.
>
> > each is listed in rcpthosts
> >
> > and a virtual domain map is put into virtualdomains
> > stooges.com:stooges
> > three.stooges.com:threestooges
> > littlerascals.com:littlerascals
> >
> > This way, there is one user for each
> > virtual domain, and the .qmail routing
> > in each of those is very specific.
> > [EMAIL PROTECTED] can have a
> > routing and [EMAIL PROTECTED]
> > can properly bounce (since shemp
> > was never a little rascal)
>
> If I only knew what you're talking about - some TV show perhaps?
>
> -----BEGIN PGP SIGNATURE-----
> Version: PGP 6.0.2 -- QDPGP 2.60
> Comment: http://community.wow.net/grt/qdpgp.html
>
> iQA/AwUBN5bPD1MwP8g7qbw/EQKMLACeMRjwDeDu7fiRs19TDzQvLBBvq14AoOvm
> 9kxh0bXzWN2W8xWrnhrhezv2
> =ZClN
> -----END PGP SIGNATURE-----
> --
> Petr Novotny, ANTEK CS
> [EMAIL PROTECTED]
> http://www.antek.cz
> PGP key ID: 0x3BA9BC3F
> -- Don't you know there ain't no devil there's just God when he's drunk.
> [Tom Waits]
>
