Re: CR exit criteria and features at risk for HTML5

On Aug 16, 2012, at 11:42 AM, Sam Ruby <rubys@intertwingly.net> wrote:

> 
> I'm strongly opposed to taking any options off the table prematurely.
> 
> Ultimately the "permissive" option will be a judgment call.
> 
> To be clear, it wasn't so far in the distant memory when IE didn't support either Canvas or SVG.  In such a context, I would agree that a statement that "we've implemented these features privately in a purely experimental build that we are not prepared to share" wouldn't be sufficient for our purposes.
> 
> On the other hand, I know of companies (such as your employer, for example), that often are reticent to prematurely release things publicly for whatever reason.  Going with that example, if there was a divergence in the spec on a minor edge case, and it was acknowledged by a representative of Apple that this was, in fact, a bug; in such a scenario I would be inclined to take a corporation with the credibility of Apple on its word.

Since you cite Apple as an example: our normal practice for CR implementation reports is that we base them on a public release of Safari, or a WebKit nightly build. I believe both of these meet the criteria of public and non-experimental. I do not anticipate that we would ever diverge from this practice.

Further, your example doesn't include use of a non-public or experimental build to judge whether interoperability requirements are satisfied, so I am not entirely sure how it is relevant. If you're imagining we would say "we've fixed this in a private branch" and use that alone to claim interop on an implementation report, I don't expect that would ever happen.


Thinking it through a little further: the "permissive" criteria suggest a qualitative judgment call on interoperability. In that context, it seems to me that a minor edge case would not even be relevant. We could have 0 browsers that match the spec on a given minor edge case and still meet that standard for the feature overall, as I understand it. Upon reflection, since the standard is qualitative judgment call and not passing a test suite, I am not really sure how we could use a non-public build to make that call.

Regards,
Maciej

Received on Thursday, 16 August 2012 19:35:45 UTC