Re: "ok" status on domains (and other objects)

[Joe Abley <jabley@isc.org>]

On Monday, Dec 2, 2002, at 14:30 Canada/Eastern, Edward Lewis wrote:

> At 18:25 +0000 12/2/02, Michael Graff wrote:
>> Once it gains RFC level, regardless of how firm the mud is, people 
>> tend to
>> point at it and say "but the RFC says..." and worse, "but the 
>> implementations
>> all do..."  The latter isn't bad, for "it's as good, but different" 
>> but is
>> IMHO a serious mistake to have for things that can be added later, 
>> once
>> they are more thought out.
>
> The problem is that the IETF has traditionally done a poor job of 
> setting expectations.  Perhaps provreg should take it upon ourselves 
> to add a note to the introduction to our documents to reinforce what 
> is said in 2026.  I don't know, I'm just saying.  The reason I think 
> that this may be applicable here is that a lot of the folks reading 
> and implementing to the documents haven't been properly warned of the 
> limitations of what an RFC is and what the standards level mean.

I don't think we can assume that TLD managers are going to read the RFC 
before they start referring to it in the "Required Compliance" sections 
of their RFPs.

In the eyes of RFP-writers, I think "RFC 8192" rolls off the tongue 
nicely and has the feeling of standardisation about it. On the other 
hand, "RFC 8192 with Michael's proposed changes to deal with the ROID 
issue, not the original idea but the one he posted after Scott 
commented on his first proposal; oh, and we want to be able to delete 
an authinfo attribute from a domain, which RFC 8192 does not allow" 
sounds decidedly non-standard and awkward.

If we wind up with twenty registries all supporting different 
variations on whatever description of EPP is first published in the RFC 
series, then we're not tremendously advanced on where we are today. 
Registrars will still need custom code to talk to different registries, 
and registries will still face expensive migration exercises whenever 
they decide they need to move from one set of registry software to 
another.

I think the spec needs to be well-done before it is published as an 
RFC. At the moment it seems medium-rare.


Joe