George Theall on 2 Aug 2004 14:58:02 -0000


[Date Prev] [Date Next] [Thread Prev] [Thread Next] [Date Index] [Thread Index]

Re: [PLUG] BIND zone oddity causing SendMail 451 Name server timeout errors?


On Sun, Aug 01, 2004 at 03:27:30PM -0400, kaze wrote:

> hostedexample.com and www.hostedexample.com both point to a real IP with no
> NAT for it setup yet. 

While it looks like you've fixed the problem (or at least I can get a
response when querying for the A record), you may be interested in the
following to explain the behaviour you were seeing:

Hostnames used as part of the SMTP MAIL FROM and RCPT TO commands are
supposed to be "canonicalized", that is, fully-qualified hostnames or
literals rather than nicknames or abbreviations (RFC 821, section 3.1). 
Yet while the RFC says canonicalization must be done, it's silent in how
to actually do it and as a result, the methods used vary from one MTA to
another, and even across versions of the same MTA.  For example, older
versions of sendmail look for records of type ANY while 8.12.0 and later
look first for A records and then either look for MX records if no A
records exist or fails if the DNS server itself fails to answer. 
Postfix, on the other hand, looks for MX records first; if there are
none or the DNS server fails *and* ignore_mx_lookup_error is yes, it
looks for A records. 

Hopefully, this explains why failure to get an A record for your hosted
domain from your DNS resulted in mail from some but not all hosts getting
through.

George
-- 
theall@tifaware.com

Attachment: pgpdX0iqzVJyi.pgp
Description: PGP signature