W3C home > Mailing lists > Public > public-web-perf@w3.org > August 2016

Re: NavigationTiming event for initial display

From: Shubhie Panicker <panicker@google.com>
Date: Wed, 10 Aug 2016 11:20:20 -0700
Message-ID: <CAK7ODi-x1JYKMZXJfCuK-TD-fnnVF=X0pFvCrgXp85yT27JacA@mail.gmail.com>
To: Stefan Seifert <nine@detonation.org>
Cc: Ben Maurer <ben.maurer@gmail.com>, Bryan McQuade <bmcquade@google.com>, public-web-perf <public-web-perf@w3.org>, Paul Irish <paulirish@google.com>
Yes measuring the timestamps for each DOM node and attaching timing info to
each node would be prohibitive performance-wise.
That said it's a reasonable question whether the developer could provide
input into "meaningfulness" of specific page renders.
However it is premature to have this discussion now, as we haven't proposed
standardizing the "first meaningful paint" (#5) yet :)
Please look out for the Explainer and github proposal over the next few
weeks, and I will also make sure that we ping this list when that proposal
is ready and out.

For this thread, it would be really helpful to get feedback on #3 First
Non-Blank Paint, and #4 First Contentful Paint.
Thanks!


On Wed, Aug 10, 2016 at 11:10 AM, Stefan Seifert <nine@detonation.org>
wrote:

> On Mittwoch, 10. August 2016 10:52:56 CEST Shubhie Panicker wrote:
>
> > The timing info is attaching to the window and not to individual DOM
> nodes.
> > As such there is no performance concern for exposing the suggested first
> > paint metrics.
>
> That's what's implemented or at least considered now. But attaching the
> information to the window brings the challenge of picking the "right"
> moment
> to capture as there can be only one number for the "First Meaningful Paint"
> metric.
>
> What I'm suggesting instead is to attach timing information to the
> individual
> DOM nodes.  As a developer I know that I'm really interested in the time
> until
> nav#main_nav is first painted and that we may completely ignore
> #splashscreen's
> appearance.
>
> And like Ben Maurer said, a single number may not cut it anyway, as there
> may
> be multiple parts of the site loaded incrementally and we would like to
> have
> timing information for them all.
>
> Kind regards,
> Stefan
>
Received on Wednesday, 10 August 2016 18:20:51 UTC

This archive was generated by hypermail 2.3.1 : Sunday, 18 September 2016 16:33:38 UTC