Re: Versioning and html[5]

Chris Wilson wrote:
> Boris said
> <>:
> I'm not suggesting you [break sites] gratuitously.  I frankly
> wouldn't expect significant numbers of pages to depend on any
> particular UA's behavior [in corner cases].  Youre incorrect in
> your expectation.

Chris, my point was that things are not black and white here.  It's not 
the case that any change you make would break a large number of sites. 
For example, if current IE behavior given a certain input is to crash, 
then it seems likely that you can change that aspect of IE behavior 
without breaking sites.

This is an extreme example, but it does demonstrate that there are 
behavior aspects which really can be changed by IE without Breaking the 
Web, by anyone's definition.  When I say "corner cases" I do mean corner 
cases.  The exact definition of what constitutes a corner case is in the 
eyes of the beholder, of course.

Please let me know if we are in fundamental disagreement at this point; 
if we are, then what I say in the next two paragraphs is not that 

If we agree so far, then in practice the question that must be asked of 
any change is how many sites would actually be affected and how much 
they would be affected.  I can accept that you consider 1% of sites as 
too many to break; I would probably agree with that, personally.  At the 
same time, you may consider a single MySpace page (or page whose purpose 
is to exploit a security hole, say) being OK to break.  Somewhere 
between those is a fuzzy band where a hard decision would need to be made.

I realize that answering the "how many sites depend on this?" question 
is difficult.  I believe Ian has been doing some research on this sort 
of question as he edited the WHATWG spec, and this working group should 
consider adopting a similar approach.

> Because I think we would eventually realize wed broken something,
> and then wed re-introduce version numbers, and the progress of HTML
> would look like this:

Just to make sure we're on the same page, is the "we" here this working 
group, or an implementor?  Or either one?


P.S.  I'm not speaking with any official hats on right now; just trying 
to understand Microsoft's position.

Received on Thursday, 12 April 2007 18:10:00 UTC