Re: [ResourceTiming] "timing allow check" steps depend on underdefined behavior

Btw, Boris, we also refer to "current document" in Navigation Timing:
http://www.w3.org/TR/navigation-timing/

Do we have the same issue there that "current document" is not defined?

Arvind


On Sun, May 25, 2014 at 7:40 PM, Arvind Jain <arvind@google.com> wrote:

> I see.
> The goal is to store timings when a navigation to a document happens. The
> timing entries are relevant for the "navigated to" document.
> So in case of navigation from about:blank to a new document within the
> same window, I think we'd want to reset the resource timings. In
> document.open() case, I think we'd want to do the same.
>
> I'm not sure how to get that behavior with the current processing model.
> Perhaps Jatinder has ideas.
>
>
>
> On Thu, May 22, 2014 at 9:37 PM, Boris Zbarsky <bzbarsky@mit.edu> wrote:
>
>> On 5/22/14, 11:25 PM, Arvind Jain wrote:
>>
>>> The current document is the as described in step 1 of the processing
>>> model.
>>>
>>
>> My point is that the term "current document" is not defined in this
>> specification anywhere I can see.
>>
>> If it's defined in some other specification, you should probably link to
>> it.  But I'm not aware of any specification that defines such a term the
>> way you want to use it here.  Same thing for "current browsing context".
>>
>> What you may want to do is to have step 1 of the processing model
>> associate a document and a browsing context with some object (the
>> PerformanceResourceTiming object?).  But then you have to watch out for
>> issues with document.open() and with initial about:blank going away but the
>> window being reused, though (those are the two cases when documents don't
>> match windows 1-1 in the platform).  I'm not sure what exact behavior you
>> want here in those situations, so I can't tell you exactly how to define
>> things.  Maybe you want to associate origins directly with some of these
>> objects instead of associating documents...
>>
>> In any case, right now the behavior is not really defined at all.
>>
>> -Boris
>>
>
>