Re: Report on testing of the link relations registry

On Wed, 1 Sep 2010, David Singer wrote:
>
> I think you are missing something very basic. If everyone does as you 
> want to do, and side-step registries that they consider are not working, 
> instead of working either to use them, or fix them, or obsolete them and 
> formally replace them, we get chaos.

This is the status quo, at least for most Web-technology-related 
registries (e.g. MIME types, link relations). I'm not suggesting that we 
ignore them. I'm suggesting we obsolete them and replace them, by 
providing something better.


> IANA is a formal registration authority with good processes, 
> trustability, and visibility.

That has not been my experience. Every interaction I've ever had with the 
IANA has been opaque, inconsistent, sometimes secretive, with no apparent 
accountability; I've had registrations dropped on the floor, take weeks to 
receive a response, get rejected for reasons I consider spurious or 
bureaucratic, and I've seen changes be made counter to processes, based 
purely on the standing of the people involved as opposed to any technical 
merit. Most people I speak to on the Web (including many Web browser 
vendor engineers) have no idea the IANA even exists, let alone that they 
are supposed to register MIME types and so on there.


> They, and those building the link relations registry there, are willing 
> to make that registry work, yes, even for you.  Either work with them, 
> or persuade IANA to dismantle this particular registry, or make 
> confusion by having multiple 'registries', some apparently uncontrolled, 
> for the same thing.

I have no problem with competition in this space. I think it would be a 
strong incentive for all involved to do a better job. People are going to 
respect the registry that provides the most accurate information and that 
is the most usable when it comes to registering new information.

Registries, like specifications, should be authoritative because they are 
the best, not because they are declared so by fiat.


> >> IANA is very successful, respected, and useful, for many many types.
> > 
> > I question all three of those premises,
> 
> "IANA is responsible for global coordination of the Internet Protocol 
> addressing systems, as well as the Autonomous System Numbers used for 
> routing Internet traffic." (quoted verbatim from their web site). How 
> well is IP addressing and routing working for you?

I'm speaking specifically about Web-technology-related registries here. I 
have no experience dealing with issues below HTTP with IANA. It's quite 
possible that IANA is excellent in that space, or maybe IANA is ignored 
in that space as it is in the Web space. I don't know. It doesn't seem 
particularly relevant to this discussion.


> > but it is not my goal to attack IANA or defend the Microformats 
> > community (or indeed the WHATWG community or the Wikipedia community; 
> > it would be equally fine by me for a registry to be in those places 
> > too). My point is just that the process for keeping track of 
> > registered link relation types should be one that is suited to today's 
> > Web, with minimal overhead. We shouldn't have to send an e-mail, wait 
> > for a human to respond, etc, when all we want to do is say "I'm going 
> > to use this type".
> 
> Well, someone should be able to check "that type is already registered 
> or in use", for a start.

The person registering the type seems like the person most interested in 
checking that information, why not have them do it? Looking up a name in a 
list is not hard... in fact it's mostly the point of a registry! :-)


> Not everyone is omniscient, that's why we have a singular registry per 
> type to keep some kind of sanity and avoid collisions.

IANA's Web-related registries (that is, the registries for things like 
link relations, HTTP headers, MIME types, URI schemes, etc) are woefully 
out of date, and no effort seems to be made to address this. They do 
nothing to avoid collisions currently.

In the case of the link relations registry specifically, the Microformats 
wiki _already_ does a better job than the link relations registry in terms 
of keeping sanity and avoiding collisions.


> Me, I am thinking of registering the domain google.co.us (which is, 
> apparently, not registered, according to whois), for myself.  Or should 
> a person check that such a registration is reasonable?

No, no person should check that. That should be (and is) automated.

-- 
Ian Hickson               U+1047E                )\._.,--....,'``.    fL
http://ln.hixie.ch/       U+263A                /,   _.. \   _\  ;`._ ,.
Things that are impossible just take longer.   `._.-(,_..'--(,_..'`-.;.'

Received on Thursday, 2 September 2010 04:26:37 UTC