Re: [Resource Timing] User feedback and a modified proposal

I'll just add to the thread that we should be very thoughtful about
how the measured times are spec'd, since some things performed by
browsers synchronously today may be performed asynchronously on a
background core in the future. For instance CSS parse time is today a
renderer-blocking event performed synchronously on the renderer core
in all browsers I'm aware of, but could be moved to a background core
in the future. In this case should CSS parse time be reported as the
amount of time spent on the background core, if none of that time was
spent blocking the renderer from making progress? Likewise IE9 parses
JS on a background core already. One could imaging decoding images on
a background core, etc.

So, what time metric do we really care about for these kinds of
events? Is it the actual number of milliseconds spent performing these
ops, or the number of milliseconds that these ops blocked the
renderer, or both, or something else? Whatever the case, we should
spec them correctly up front, so that as browsers evolve the spec'd
definitions still hold.

-bryan


On Fri, Jun 3, 2011 at 3:09 AM, James Robinson <jamesr@google.com> wrote:
> The -beta- idea is just a bad idea, as was made clear in the discussion on
> the blog post suggesting it. I don't see any reason to spend time worrying
> about it at all.
>
> - James
>
> On Jun 3, 2011 12:04 AM, "Sigbjørn Vik" <sigbjorn@opera.com> wrote:
>> On Wed, 01 Jun 2011 19:40:59 +0200, Jatinder Mann <jmann@microsoft.com>
>> wrote:
>>
>>>> I take my words back regarding the prefix. The new vendor prefix is
>>>> -beta-, one prefix to rule them all. See
>>>> http://www.quirksmode.org/blog/archives/2010/03/css_vendor_pref_1.html
>>>> Should we be avant-garde here, and be the first to implement -beta-?
>>>> (I don't think any other specifications have picked up on this to
>>>> date.)
>>>
>>> What would the expected behavior be if two user agents use the same
>>> initiator type name but they mean something slightly different? It would
>>> seem that having a vendor prefix, instead of a beta prefix, might be
>>> helpful in distinguishing the two.
>>
>> I don't see that being the case. If user agent A has implemented
>> -beta-compileTime, user agent B would probably choose to call their
>> implementation something else if it is not compatible, for instance
>> -beta-compilationTime or -beta-ESToMachineCode. -beta- allows multiple
>> vendors to use the same name, but it doesn't make it a requirement.
>>
>> I don't have any strong opinions on this, but I recalled the discussion,
>> and the conclusion I linked to seems to make sense to me. I see little
>> need to have that same discussion again on this mailing list, so as long
>> as an informed decision is made, I'll be happy regardless of which
>> conclusion is reached.
>>
>> --
>> Sigbjørn Vik
>> Quality Assurance
>> Opera Software
>>
>

Received on Friday, 3 June 2011 15:29:34 UTC