home *** CD-ROM | disk | FTP | other *** search
- Xref: sparky comp.mail.misc:3774 comp.mail.uucp:2121 comp.mail.sendmail:2854
- Newsgroups: comp.mail.misc,comp.mail.uucp,comp.mail.sendmail
- Path: sparky!uunet!utcsri!geac!torag!robohack!feline!becker!bdb
- From: bdb@becker.GTS.ORG (Bruce Becker)
- Subject: Re: dxmail on ULTRIX
- Message-ID: <1992Nov23.010740.18068@becker.GTS.ORG>
- Organization: G. T. S., Toronto, Ontario
- References: <1992Nov12.182035.24694@decuac.dec.com> <1e15ghINN2oi@ef2007.efhd.ford.com> <1992Nov16.142027.21114@decuac.dec.com>
- Date: Mon, 23 Nov 1992 01:07:40 GMT
- Lines: 86
-
- In article <1992Nov16.142027.21114@decuac.dec.com> avolio@dco.DEC.COM (Frederick M. Avolio) writes:
- |
- |If I have an address
- |
- | something!else%host.UUCP@domain
- |
- |while it is ugly, and I'd rather die than generate such an address, the
- |assumption is that "domain" will handle what is on the right hand side and
- |if it generated it, it must be able to handle it.
- |
- |While all of this is ugly, it is allowed because it is on the
- |left of the @ sign. There is only one way to handle the above address.
- |Send it to "domain" for "domain" to handle "something!else%host.UUCP" however
- |it wants to. It is not a question of defined or undefined. I have no
- |business looking at what is clearly the localpart of the address.
- |
- |The case in question was a sendmail.cf that erroneously looked at this address
- |and decided it should go to "something". This is wrong. The sendmail.cf
- |in question actually took the address (again, using the above example) and
- |changed it to
- |
- | something!else%host.UUCP<@domain>
- |
- |correctly focusing on "domain" and then decided, no, there is a bang in
- |there so then it changed it to the erroneous:
- |
- | else%host.UUCP@domain<@something.UUCP>
- |
- |Based on that, it sent it via the wrong transport to the wrong host. The
- |corrected sendmail.cf that I sent fixed this problem.
- |
- |The point is that anything to the left of the @ is up to the host to the
- |right of the @ to handle. You -- and the ultrix sendmail.cf shipped with
- |the product -- are breaking the rules. You are peaking. No peaking allowed.
-
-
- Yes, the sendmail.cf in Ultrix is broken,
- in more ways than just this. So also are
- most that are shipped - a visible example
- of what else may also be wrong if one was
- to look at the source for many things. A
- way of fixing this problem is needed, but
- there seems to be difficulties deciding a
- proper course, to wit:
-
-
- Sadly the interpretations in the uucp and
- the internet worlds differ. Mailers such
- as smail 2.5 for uucp sites understand no
- special interpretation for "%", so in the
- case above delivery fails for "something"
- to send to local user "else%host.UUCP" (a
- patch exists which translates "%" into an
- "@" and attempts to resolve the result).
-
-
- "something!else%host.UUCP@domain" thus is
- routed "domain!something!else%host.UUCP".
- The mailer at "something" hopefully would
- route "host!else", if it's able to.
-
-
-
- In the domain system only the domain name
- is relevant for routing purposes, but the
- situation is not so well defined when the
- target system's mailer is invoked. There
- seems to be a belief that "%"'s should be
- resolved as soon as they are seen, rather
- than by the mailer of last resort as it's
- usually practiced in the uucp world.
-
-
- "something!else%host.UUCP@domain" thus is
- "something!else@host.UUCP" via "domain" -
- this is clearly not at all the same as in
- the uucp case; there's no simple means to
- resolve it, since this seems to be a kind
- of "religious" difference...
-
-
- --
- ,u, Bruce Becker Toronto, Ontario
- a /i/ Internet: bdb@becker.gts.org Uucp: ...!web!becker!bdb
- `\o\-e
- _< /_ DO NOT TAUNT HAPPY FUN BUSH - H. R. Perotnoid
-