Moving past last call for HTML5

On Thu, 19 Feb 2009, rubys wrote:
> 
> You've made amazing progress to date, and I very much want you to 
> continue.  However from my perspective you have done so by cutting a 
> number of big corners: including the need to establish and maintain 
> consensus, and the need to create test suites up front.
> 
> Frankly, the first corner I mentioned will bite you in the ass when we 
> get to the very next milestone.  I haven't reviewed the list of things 
> that Rob has cut from your draft to produce his draft, but I'm confident 
> that a number of them are things over which consensus has not been 
> reached.  In fact, many of them appear to be outside of the scope of the 
> W3C Working Group.  Could the charter of the working group be changed 
> and consensus be reached on any one of those items? Quite possibly.  
> Could this be done simultaneously for each and every one of the items 
> and on the schedule you have defined?  Not a chance.

Realistically speaking, we'll never have complete consensus on everything 
in HTML5. At the simplest level, there are contradictory opinions on the 
very fundamentals of the work -- some people want error handling defined, 
some don't; some people want a schema, some don't; some people want APIs 
defined, some don't; the list is long.

So consensus -- unanimity -- isn't an interesting goal.

The next step down in terms of opinion-based progress is majority 
agreement, and I am confident that with the exception of things that need 
changing and will be changed in time for the next milestone, we have a 
majority agreement on everything in HTML5.

Majority agreement in a self-selected community like an open working 
group is worth less than it would appear, though, because there is a 
selection bias: only people who are interested in both the technology and 
in standards development are going to take part. In the W3C working group, 
there is a further bar: we only allow people who are willing to put up 
with an inordinate amount of bureacuracy (to join) and noise to be part of 
the group whose opinion is measured.

Statistically, therefore, the opinions of the working group almost 
certainly don't match the opinions of the whole Web community.

An alternative approach, the one which the WHATWG (and by extension the 
editor's draft in the HTMLWG) has been using, is to use the "vote with 
your feet" approach: things that aren't implemented or used get thrown 
out, things that _are_ implemented or used get adopted. For new features 
this means that browser vendors have, as a group, the ability to veto 
anything, but that's true anyway, whether we admit it or not. For the 
direction the Web is going in, it means that authors that write content 
that gets crawled and measured in the studies we do get a vote. In this 
model, users get the ultimate vote by deciding whether or not they adopt a 
particular browser and whether or not they visit particular sites; 
browsers without adoption end up with less ability to veto, and sites with 
fewer users tend to die earlier and fall off the statistical radar.

I understand that it would be controversial to use this method in the W3C, 
however.

Given the goal of publishing a Last Call draft in October, what approach 
do you think we should take to achieve that goal in the W3C? Should we 
start having votes to ask people which sections they won't support? Should 
we start collecting "formal objection"-type comments on the list? What bar 
will you be looking for when I say "ok there is no outstanding feedback, 
let's publish a Last Call draft"?

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

Received on Thursday, 19 February 2009 21:52:40 UTC