Re: [dnsext] Obsoleting SPF RRTYPE

[Masataka Ohta <[e-mail]>]

Andrew Sullivan wrote:

>> implication of this would appear to be that definition of new RRs
>> should be constrained to "consenting adults" (e.g., EUI48/68). Is
>> there some other takeaway from the "failure" of the SPF RR?
> 
> One obvious one is, "Don't have two mechanisms to achieve the same
> goal."

It means anything to be implemented by TXT can not have its own
RR.

> Given that SPF has the TXT record and that there is zero hope
> of getting rid of that use, the SPF RRTYPE was just a vain hope.

"there is zero hope of getting rid of that use"?

Why?

Some reasoning, please.

> And indeed, as DKIM and other such uses show, the underscore-label
> extension is a permanent part of the DNS landscape and we should
> expect applications to depend on that as often as on new RRTYPEs,
> regardless of any advice to the contrary.

Where is the advice not to have underscore-label extensions
with SRV?

The problem of underscore-label extensions is that the
underscore-labels is not assured to be unique (RFC1034
imposes no restriction to use underscore-label for any
purpose) and can have conflicts to other applications.

Thus, an application specific underscore-label extension is
just fine if and only if it is tied with applications' own
specific RR type to prevent ambiguities on meaning of the label.

That is, having application specific RRs is MUST, even if you
use underscore-labels.

						Masataka Ohta