Re: [ietf-provreg] References for Today's Host Object Discussion

[Edward Lewis <edlewis@arin.net>]

At 8:59 +0200 4/11/03, Jörg Bauer/Denic wrote:

>The idea behind it is that it doesn´t forbit you to just use NS and (glue)
>A Records, it´s up to the policy, but it also alows you to store other
>records into the zone.
>Just think about DS Records (for DNSSEC). Maybe it makes sence to get
>these records too.

Under the banner of getting the first job done first without 
squelching improvements...

DS RR's are still not defined in any archived document (e.g., RFC), 
they are not even experimental, much less standards track, so I'd 
hesitate to think that far ahead.  Yes, doing whatever is necessary 
to facilitate DNSSEC is something I've been pushing for for quite 
some time, but it's not appropriate to tie EPP (version 1.0) to that 
boat anchor at this time.

Scott Hollenbeck has written a non-WG draft describing how DNSSEC 
could be accommodated in EPP extensions.  Note that this is a live 
debate, but just not one being conducted in the working group.  Nor 
is the debate, although alive, progressing anywhere rapidly, so far 
as I know, it's just there.

>Of cause you can develop extensions for all this kind of stuff, but i see
>a BIG problem with interoperability if al lot of Registries are going to
>have their own extensions.

This is what we are trying to avoid.  But sometimes it is hard to 
anticipate what the greatest common denominator is before folks run 
off and do (hopefully prototypes of) non-interoperable things.

I understand the dilemma, but the IETF process is built on putting 
engineering before the realities of business needs to maintain a 
revenue stream.  I'm not saying that the IETF process is necessarily 
correct in this manner, just that there is a conflict here.  This is 
what makes the job of chairing a group like this, umm, so rewarding. 
;)

>I also see the  drawback of this approach: the records aren´t "well
>defined". You have a name, a type and all the rest. But to have some more
>level of flexibility inside the protocoll, i prefer also to have such kind
>of mapping.
>
>Maybe we can put two type of mapping inside:
>1. a pure DNS+Glue mapping
>2. a "whatever you want, put into DNS" mapping

Number 1 is a pretty clearly defined case.  Number 2 is not, and 
dealing with problems that are not clearly defined is a recipe for 
disaster when you are expecting to meet a deadline or a near-term 
goal of reaching a basic protocol.

-- 
-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-
Edward Lewis                                            +1-703-227-9854
ARIN Research Engineer

   ...as rare as a fire at a sushi bar...