Re: Resource Timing Level 1 and beyond

On Tue, Apr 26, 2016 at 7:52 AM, Philippe Le Hegaret <plh@w3.org> wrote:

> Following last week discussion, I added "Level 1" to Resource Timing with
> the following:
> [[
> This specification is ready for wide review, with the following features
> at risk for the first release:
> *    Dependency with Performance Timeline 2, since performance observers
> are lacking implementations;
> *    Dependency with High Resolution Time 2 and workers support, including
> workerStart, since we're still refining time origin;
> *    nextHopProtocol, transferSize, encodedBodySize, and decodedBodySize,
> since we're currently lacking implementations.
> ]]
>

We also had secureConnectionStart marked as optional for a long time and
recently changed it to mandatory. My proposal would be to also treat that
change as an L2 feature. With these carveouts in place, I think we should
have three existing implementations (Edge, FF, Chrome) of proposed L1. And
once we land https://github.com/w3c/resource-timing/issues/46, we can
(hopefully :)) confirm that.


> Imho, the issue that affects the most implementations at the moment is
>  https://github.com/w3c/resource-timing/issues/12
> I'm proposing that we don't solve it for V1 but keep flagging it as an
> issue in the spec for Web developers to be aware of.
>

I agree. The spec did not indicate either way until we landed [1] and I
think we can: (a) keep it as such for L1, (b) resolve it in L2. With that
in mind, we would probably need to back out that commit for L1?

[1]
https://github.com/w3c/resource-timing/pull/19/commits/0eb0f6997fc3f8a70a556212b45fa9ce5cfe7631

If we're ok with this, plan is to move a Level 1 version of the spec
> without the feature at risk and publish at the same time a Level 2 of spec
> as normal. Level 1 shouldn't impact editors, ongoing issues, or pull
> requests. The branch gh-pages will continue to hold v.next and
> https://www.w3.org/TR/resource-timing will continue to reflect it as a
> Working Draft. In other words, Level 1 is and should remain a side
> artifact. We do however have enough implementations of Level 1 to ship to
> Recommendation within 3/4 months.


sgtm. Much overdue, but better late than never! :)

ig