- From: Mark Nottingham <mnot@mnot.net>
- Date: Thu, 18 Feb 2010 10:01:48 +1100
- To: Maciej Stachowiak <mjs@apple.com>
- Cc: Julian Reschke <julian.reschke@gmx.de>, Ian Hickson <ian@hixie.ch>, Paul Cotton <Paul.Cotton@microsoft.com>, "public-html@w3.org" <public-html@w3.org>
On 18/02/2010, at 12:24 AM, Maciej Stachowiak wrote: > Like I said, I'm not going to object on this basis, I was just curious to hear from Mark why there is not any form of provisional or experimental registration. I think you can bucket "provisional registrations" as: 1) reserving names for specs that aren't published yet 2) just too freaking lazy to write the spec 3) proprietary extensions 4) collecting named used 'in the wild' to prevent conflicts #1 is already accommodated in the link draft; see the language about allowing pre-registration if the DE is satisfied that a spec is forthcoming. #2, IME (e.g., with the HTTP header registry) doesn't happen that often; i.e., someone who is so lazy to write a spec, they're probably too lazy to register something. I.e., it automatically gets turned into #4 if they get any deployment, and it's not a problem if they don't. #3 is accommodated by extension relations. #4 is the case that I think you're concerned about here. What I've observed is that just allowing provisional registrations wholesale encourages #1-3 to use the provisional registry and stop there -- thereby increasing the number of relations out there without proper specification, review, etc. So, rather than encourage that kind of behaviour, what I'd rather see is allowing lightweight specs to be written which document (sometimes very briefly, with appropriate warnings) existing, widely-used link relations (NOT speculative ones; i.e., the #4 case, and only that case) to get them into the registry, possibly with a note to the effect that they're provisional (or some similar state). The main differentiator here is that going through the "normal" process results in a relation being minted, specified and registered by the same party, whereas here we're talking about a third party registering it, when it becomes clear (in the judgement of the DE, ultimately) that a) it's had wide deployment, and b) that entity isn't going to register it in a reasonable amount of time. I think this fits into the existing process, although it would require some interpretation by the DE. I'd be happy to document it if that makes people more comfortable. Thoughts? -- Mark Nottingham http://www.mnot.net/
Received on Wednesday, 17 February 2010 23:02:21 UTC