Re: http content type authoritative for object data?

Jonas Sicking wrote:
> ...
> I should add that it's not entirely impossible that this could be made
> to work, with sufficient evangelism efforts. Bug 1156 [1] (one of few
> bug numbers I know by heart because it caused so many regressions)
> changed things to match the HTML4 spec. We were able to stay with that
> behavior fairly far into the Firefox 3.0 release before we were forced
> to revert behavior in bug 395110 [2].
> 
> It is however possible that with the right combination of logic and
> the right amount of evangelism effort to get a few high profile sites
> fixed, it's possible that the http mime type could override the @type
> mime type. I'm a bit reluctant to take a lead here though given our
> previous failed attempt.
> 
> [1] https://bugzilla.mozilla.org/show_bug.cgi?id=1156
> [2] https://bugzilla.mozilla.org/show_bug.cgi?id=395110
> ...

Summarizing from my point of view:

- HTML5 currently requires ignoring authoritative metadata in some 
cases. The information in [2] indicates that in at least one 
measurement, this "fixed" around 1% of pages using <object>

- Firefox changed to the HTML5 behavior (or something close to it) early 
2008

I'm not convinced that the behavior required by HTML5 should be required 
  for UAs. It's certainly incompatible with the specified behavior in 
HTML4, and the general web architecture. So minimally, UAs should be 
*allowed* to respect the metadata sent with the referenced resource's 
entity.

Best regards, Julian

Received on Wednesday, 9 December 2009 08:54:15 UTC