Re: [Arcing] New Version Notification for draft-stw-whatsinaname-00.txt

[Suzanne Woolf <[e-mail]>]

On Jul 12, 2016, at 5:47 PM, "John R Levine" <[e-mail]> wrote:

>> Probably more importantly, I don't think "there's nothing we can do" is a good answer for protocol developers within the IETF.
> 
> Oh, I agree, but if we're just giving advice, wouldn't that be about two sentences?  If you want to use a domain name, register one and use that as the root of your tree, or else use something in *.alt.

That draft would provide no guidance at all on deciding whether you want to use a domain name, use something from the subset of domain names that are DNS names, or do something else. 

It would also provide no guidance at all on the characteristics of an acceptable string. Just to take an easy example, why on earth would you want to risk collision by choosing a human-friendly string as a single label DNS name when your protocol is strictly a locally-scoped machine-to-machine variant of DNS, and no human will see the names it uses in the course of normal operation? "I want to reserve my cat's name as a single label" is a nice-to-have in that case, not a need-to-have, and probably a thing to avoid, particularly if it comes with the documented certainty of collision with a name already in the global-scope DNS. (Or not-- people use split-horizon configuration in their DNS all the time, committing to different answers for "the same name" depending on the implicit scope of the query. But it works a lot better if they do it on purpose.)

"If you want to use a domain name" is an enormous "If", and an answer comes at the tail end of a process that's currently driven primarily by whim and mythology. Maybe we can do better.


Suzanne