On Friday, February 24, 2012 12:57:49 PM Andrew Sullivan wrote:
On Fri, Feb 24, 2012 at 12:47:10PM -0500, Scott Kitterman wrote:
It's occured to me that it might be useful to pre-allocate some new
types without a current use assigned (e.g. TXT1, TXT2, TXT3) so that
there's time for them to be integrated into tools before they are
How could you integrate a pre-allocated type when you don't know what
its format is? If you knew how to do that, you could just deal with
unknown RRTPYEs, and then having TYPE99 wouldn't be a problem anyway.
If there had been a TXT1 ... N in 2004, SPF (to put an example) could have
picked TXT1 (assuming it wasn't used by something else). Then later the label
could have been changed to SPF once usage was established and standardized.
Then a few years later, DomainKeys could have used TXT2 (or whatever) for it's
key records. As it was, a new rr type wasn't (AFAICT) given serious
I'd give them all a format of text and if a protocol needs something more
specific than that it can interpret it at the application level.
protocol to get deployment it needs the new type assigned. For a
new type to get assigned, there needs to be some indication it will
Absolutely false. RRTYPE assignment is by expert review, and the last
reviews, while late and a little careless about process, were not in
any way held up by the possibility (or lack thereof) of use. Nobody
is reluctant to assign RRTPYEs. We have lots of numbers. There may
have been problems with this in the past, but not at least since I've
been DNSEXT WG co-chair, which is several years now.
OK. My one experience with this was the assignment of Type 99 as Type SPF.
It wasn't done until very shortly before RFC 4408 was published (and by then
it was far too late). If it's different now, then my idea my well be obsolete.
type there's no incentive to migrate. The only way to break this
cycle is to have something readily available for initial use with
There's lots of room for that, too. The range 65280-65534 is for
private use. You can do whatever you like there.
Yes, but that still assumes a transition to a new rr type when the protocol
matures. The one experience that we have with trying to do this (that I know
of) doesn't inspire confidence that this is a reasonable expectation.
Ietf mailing list