Re: Registry policies

On 01/03/2012, at 7:24 PM, Julian Reschke wrote:
>> I'm having a hard time believing that Cache Directives, Range Specifiers and Warn-codes should be IETF review. How do people feel about making them Specification Required?
> 
> Warn codes are scarce. We already burned lots of cycles on a registry that's probably never be used; so I'm -1 on any more changes on it.

Well, we burned a lot of cycles *talking* about it; actually doing it was trivial.


> Cache Directives and Range Specs: maybe. They are they way they are because we used IETF Review as default.
> 
>> Does FCFS really make sense for upgrade tokens? It seems like this should be Specification Required, at a minimum. Yes, I know that it's historically been FCFS, but we have the latitude to review registration policies.
> 
> If we're ready to change existing procedures, then yes, we should change this.
> 
>> Finally, all of the Specification Required registries (including any we decide to convert) imply use of an expert reviewer; we should make sure that we give reviewers advice.
> 
> And we should consider how well Expert Review works in practice (ahem!), and how many DEs we'll be able to recruit. Maybe this is an argument in favor of IETF Review.


While I agree that ER is tricky to get right, the problem with IETF Review is that the registry gets far away from practice. I'd very much like to be able to document values that are in use.

Thinking out loud, IETF Review would be ok if we augmented it with a DE-like function -- call it a Registry Maintainer -- whose job would be to submit other values that are in use and have them designated with a special status, so that a) the registry was (more) complete, and b) collisions were avoided.


--
Mark Nottingham   http://www.mnot.net/

Received on Friday, 2 March 2012 00:03:02 UTC