Re: #346: Registry policies

On 28/03/2012, at 2:14 PM, Roy T. Fielding wrote:

> On Mar 28, 2012, at 1:27 PM, Mark Nottingham wrote:
> 
>> <http://trac.tools.ietf.org/wg/httpbis/trac/ticket/346>
>> 
>> At yesterday's meeting, there was some pushback on defining our registry policies as IETF Review for "consistency."
>> 
>> Given that there is a larger discussion about registry policy definition taking place, and I *think* we've agreed that we shouldn't block HTTPbis on that discussion (since it's likely to take some time), it seems that we should take the attitude of installing (relatively) temporary registration policies; i.e., we should make them reasonable, but not try to solve all of the problems we perceive with them, in the belief / hope that a more general effort will help later.
>> 
>> In #346, we changed the following registries to "IETF Review":
>> 
>> * Upgrade Tokens (previously First Come First Served)
>> * Transfer-Codings (previously Specification Required)
>> * Content-Codings (previously Specification Required)
>> 
>> I'm re-opening this ticket based upon discussion in the meeting yesterday. 
>> 
>> My take - 
>> 
>> I believe we should leave Transfer-Codings and Content-Codings as IETF Review, because otherwise we will need to establish expert review procedures and guidelines for them, as well as identify experts. These are very low-throughput registries, and will benefit from IETF review (as there's a cost to adding new schemes to negotiation).
> 
> Why do we care?  FCFS is fine provided that reservations for deployed names
> can be overtaken by the controllers of those names.  What I'd like to see in
> a registry is a status section that is subject to expert review, so that I
> can mark registered names as "standard" / "experiment" / "deprecated" / "idiotic".

RIght; I want to get there too, but I don't think we should block on figuring that out. See other message.


> 
>> I think we should discuss Upgrade Tokens; first-come-first-served may make sense here. However, I'd note it'd be a shame if we spent too much time on it.
> 
> FCFS is fine provided that reservations for deployed names can be overtaken
> by the controllers of those names.
> 
> ....Roy

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

Received on Wednesday, 28 March 2012 15:15:57 UTC