[dns-operations] MX record definition?

[Mark Andrews]

In message <20081218142445.GA28190 at vacation.karoshi.com.>, bmanning at vacation.ka
roshi.com writes:
> On Thu, Dec 18, 2008 at 05:57:31AM +0000, Paul Vixie wrote:
> > > > 	Seeing a IP address in the exchange field of a MX record
> > > > 	in a master file is, almost always, a configuration error.
> > 
> > in that it will never work reliably and is rarely what's intended,
> > that's a literally true statement, but in no way germane to the topic.
> 
> 	as you point out below - none of this is germane to the
> 	format of an MX - the DNS is a bearer channel.  
>  
> > > > 	Seeing a IP address with a period at the end in the exchange
> > > > 	field of a MX record in a master file is also, almost always,
> > > > 	a configuration error
> > 
> > even moreso than the above, on all counts.
> > 
> > > 	sounds to me like -in some cases- a fully numeric string that 
> > > 	-could- be parsed as an IP address would be legal.
> > 
> > now you're changing the subject from whether it's a config error to
> > whether it's legal.  it's always legal, but can be meaningless in a
> > way that does not look meaningless, and therefore always dangerous.
> 
> 	if something is legal then why would it be a configuration error?
> 
> 
> > > > 	MTA that treat a "123.321.23.123." as a IP address are broken.
> > > > 
> > > > > As far as being a no-no: they aren't prohibited by the DNS 
> > > > > specification.  Perhaps there's a mail protocol restriction.
> > 
> > DNS is just a bearer channel for this.  RFC 974, RFC 1123, and others
> > talk about the content and meaning of the MX RDATA.  an MTA who treats
> > the "123.321.23.123." domain in an MX RDATA as an IP address isn't so
> > much broken as it is acting outside of the internet standards.  the
> > reason that such MX RDATAs are consider configuration errors is that
> > since treatment of same as an IP address cannot be depended upon and
> > is somewhat rare.
> 
> 	if an MTA treats 123.321.23.123. as an IP address, i'd claim 
> 	sloppy parsing...  but then i've always been in favor of FQDNs.
> 	123.321.23.123 would be an IP address and if used where it
> 	would be parsed as a domain name, we are stuck w/ the rule on
> 	appending the domain suffix.   which brings in search rules again.

	A MTA which applies a search list to EXCHANGE is broken.
 	
> 	try:
> 
> 	123.321.23.123.gen
> 	123.321.23.123.gen.ca		<---  this might be fun
> 	123.321.23.123.gen.ca.us.
> 
> --bill
-- 
Mark Andrews, ISC
1 Seymour St., Dundas Valley, NSW 2117, Australia
PHONE: +61 2 9871 4742                 INTERNET: Mark_Andrews at isc.org