Re: [mif] [dnsext] [DNSOP] 2nd Last Call for MIF DNS server selection document

[Keith Moore <[e-mail]>]

On Oct 23, 2011, at 2:39 AM, Matthew Pounsett wrote:

> 
> On 2011/10/22, at 15:21, Keith Moore wrote:
> 
>> 
>> On Oct 22, 2011, at 2:42 PM, Doug Barton wrote:
>> 
>>> 1. I think we're all in agreement that dot-terminated names (e.g.,
>>> example.) should not be subject to search lists. I personally don't have
>>> any problems with any document mentioning that this is the expected
>>> behavior.
>> 
>> agree.  however there are standard protocols for which a trailing dot in a domain name is a syntax error.
> 
> Any protocol that makes a standard FQDN a syntax error is itself in error.  Not to say that these don't exist, but if people are writing protocols that can't deal with a properly formatted FQDN they need to stop.  Now.

Per RFC 952, a standard FQDN does not contain a trailing dot.   Neither do email addresses nor domains in URLs.    Changing that set of embedded practices is much more difficult than changing the expectations of the relative few who currently expect names with dots to be subject to search lists.

>> Strongly disagree.  That would leave users without a protocol-independent way of unambiguously specifying "this is a fully-qualified domain name".
>> 
>> The practice of applying search lists to names with "."s in them needs to die.
> 
> I can't agree with this statement.  As others have said, the practice of using a search list to allow 'ssh foo.bar' to reach 'foo.bar.example.com' isn't going anywhere, and there are a lot of people that make extensive use of the convenience.

It needs to die because it's fundamentally broken.   Vanity TLDs will only make it worse.   I understand that there are sites that use it and people who are accustomed to it.   I don't pretend that we can stop them.   We can, however, explain the negative consequences of doing this (some of which might be specific to systems with multiple interfaces), and recommend that they transition away from that practice.   And recommendations for systems with multiple interfaces can be chosen in such a way as to allow search lists to break even more.

Keith