On Fri, Aug 30, 2013 at 9:35 AM, John C Klensin <john-ietf@xxxxxxx> wrote:
-- Hi.
Inspired by part of the SPF discussion but separate from it,
Patrik, Andrew, and I discovered a shortage of registries for
assorted DNS RDATA elements. We have posted a draft to
establish one for TXT RDATA. If this requires significant
discussion, we seek guidance from relevant ADs as to where they
would like that discussion to occur.
Three notes:
* As the draft indicates, while RFC 5507 and other documents
explain why subtypes are usually a bad idea, the registry
definition tries to be fairly neutral on the subject -- the idea
is to identify and register what is being done, not to pass
judgment
RFC 5507 does indeed say that but it is an IAB document, not an IETF consensus document and it is wrong.
The document was written by authors who wanted to promote the notion that the DNS should accept deployment of new RR types and to do otherwise is a standards violation. So it makes claims about some subtyping strategies that are wildly inaccurate.
Prefixing works fine unless you need to use wildcards. What the draft does not mention is that it is easy to fix the wildcarding problem. (see XPTR).
The consequence of this is that we still don't seem to have a registry for DNS prefixes, or at least not in the place I expect it which is
Domain Name System (DNS) Parameters
Instead we have a bunch of registries that are protocol specific. Which when the prefixes are from a common namespace is borked.
The IANA should be tracking SRV prefix allocations and DNSEXT seems to have discussed numerous proposals. I have written some myself. But I can't find evidence of one and we certainly have not updated SRV etc. to state that the registry should be used.
Fixing TXT is optional, fixing the use of prefixes and having a proper registry that is first come first served is essential. Right now we have thousands of undocumented ad-hoc definitions.
This task has to be taken away from people who think cutting a new RR is always the answer. It isn't. We are going to have hundreds of thousands of Web Services and the DNS RR space is a 16 bit number.
Service discovery requires prefixes.
Here is a draft that works fine (except for the IETF review mistake). Just put IETF last call on it:
Website: http://hallambaker.com/