[dns-operations] MX record definition?

Eric Brunner-Williams brunner at nic-naa.net
Thu Dec 18 14:58:45 UTC 2008

Edward Lewis wrote:
> At 17:08 +1100 12/18/08, Mark Andrews wrote:
>>     Which would only be legal if you are running in your own private
>>     namespace.
> Bingo!
>>     Not on the big I Internet.  Maybe in your own private namespace.
> The DNS is not only for the global public Internet.
> Now, to get back to the original question - this thread is a reason 
> why there is no document saying it is really poor practice to put an 
> all numeric domain name in an MX record.  On the one hand, it is true 
> that it won't work on the global public Internet because all numeric 
> domain names are not defined and never will be.
> (We can't have an all numeric TLD under the global public Internet's 
> root zone as managed by IANA/ICANN because of IDN issues with the 
> directionality of characters.  Punctuation ['.'] and numbers [0-9] 
> have no directionality in the character hence you can't determine if a 
> portion of a domain name is localized as ".0" or "0."  Yadda, yadda, 
> yadda, IOW, numbers might 'jump' over labels.  See 
> http://stupid.domain.name/node/683 for a start on that.)
Hi Ed,

I appreciate where you're going with this but there are some nuances. 
The "no digits (octal, decimal or hex representations) rule" in the 
current Guidebook for Applicants is motivated by our old friend the 
in_addr routines, and I've sent note in to correct the next version of 
the draft. There is a separate concern that because of the Unicode bidi 
algorithm treating "." as punctuation, and not as a label separator, 
that the weak directionality of punctuation allows directionality to 
leak from one label to the next, the point of the example at PAF's site 

This note could be much longer.
> So, numeric names in MX records is a bad idea (on the global public 
> Internet).  But the reason it isn't documented (within DNS) is that it 
> is legal to do so as far as the protocol is concerned and can be made 
> to work in some environments.  Whenever someone tries to argue that 
> "it's legal in the protocol" someone else responds with "but it won't 
> work" and we wind up seeing about a half-dozen or so e-mails fighting 
> it out, followed by a lull of a few months until the next question 
> comes up on the list.

I still fail to see the harm of pi being a text record, for any 
approximation of pi, for any text record. I go to 3.14... (63 digits) 
when I want the first million digits of pi.
> Merry festivus (http://en.wikipedia.org/wiki/Festivus) to all...


More information about the dns-operations mailing list