MX question

Chris Thompson cet1 at hermes.cam.ac.uk
Wed Aug 23 14:15:46 UTC 2006


On Aug 22 2006, Barry Margolin wrote:

>In article <ecf9l4$cjq$1 at sf1.isc.org>, "Jim" <jim at compton.net> wrote:
>
>> hi folks,
>> 
>> I have a customer's dns server (bind - not sure of version) not able to 
>> resolve my mx records to one of my zones. Here are my current records:
>> 
>>  100 mydomain.com.s5a1.psmtp.com.
>>  200 mydomain.com.s5a2.psmtp.com.
>> 
>> I'm being told from the customer that my mx records are formatted 
>> improperly. His argument is that his version of bind is unable resolve the 
>> mx record when the domain name _does not_ appear at the end of the record 
>> (ie s5a1.psmtp.com.mydomain.com would work fine for him).
>> 
>> Is this is a known problem with some versions of bind?
>> Can someone point me to documentation that tells me proper/illegal syntax 
>> for my mx records.
>
>MX records are allowed to point to any name that has an A record.  Is he 
>claiming that the MX record has to point to a name in the same domain?
>There has never been such a requirement, and there are many domains that 
>do not fit that criterion.

On the assumption that the OP is not obfuscating his example too much ...

mydomain.com.s5a1.psmtp.com and mydomain.com.s5a2.psmtp.com do resolve
to A records for 64.18.4.10 and 64.18.4.11 respectively. But they do so
by virtue of wildcard A records, and reverse lookup of the IP addresses
yields PTR records containing "*.s5a1.psmtp.com" and "*.s5a2.psmtp.com".

There's nothing technically invalid about that[*], but it could well
upset an MTA using a sufficiently paranoid resolver.

[*] ... or is there? The target of an MX is meant to be a canonical name.
not (e.g.) a CNAME. Does using a wildcard address record mean that it
isn't <<really>> a canonical name?

In any case, I don't see the point. Using different aliases of an address
in MX records for different mail domains isn't going to get you anything
useful at the SMTP level, AFAICS.

-- 
Chris Thompson
Email: cet1 at cam.ac.uk



More information about the bind-users mailing list