Re: [NavigationTiming] RFC: ticking navigationStart

I'd like to see navigationStart as a consistent mark of the time that the
user committed the navigation (just like it is in the Chrome and IE
implementations). Then the application can count on it as the beginning of
user-perceived latency. If desired, the application can use any other
details such as redirect time to filter out phases that they don't care
about.

That being said, if exposing navigationStart as currently implemented leads
to new privacy concerns, then we must come up with a contingency plan.
According to the Chrome security/privacy team's findings there are no new
privacy issues with navigationStart [1]. However, I haven't seen a response
from IE and FF on that thread yet. If a novel privacy exploit is
demonstrated, then I'd like to explore whether removing/guarding other
phases such as redirect and/or unload would allow us to keep navigationStart
as-is rather than jumping to mucking with navigationStart. Maybe a novel
privacy exploit has been demonstrated and I missed it; if so could we lay it
out on this list?

-Tony

[1] http://lists.w3.org/Archives/Public/public-web-perf/2010Oct/0057.html

On Tue, Oct 26, 2010 at 1:02 AM, Zhiheng Wang <zhihengw@google.com> wrote:

>   In last week's conf call, there was a discussion about the appropriate
> spot to time navigationStart in case of a redirect.
> To elicit feedback from more audience, here is a summary:
>
>       |-------redirect---------|----------destionation----------|
>      A                          B                                    C
>
>   * Proposal 1: in case of a redirect, navigationStart points to the start
> of the destination page, i.e., Point B in the above
>     diagram.
>     - navigationStart is mostly isolated from the effect of unloading the
> previous page so it doesn't have be guarded by the
>       same origin policy like redirectStart/End and unloadEventStart/End.
>     - navigationStart will not be the starting point of user experience in
> case of a redirect, but redirectStart.
>
>   * Proposal 2: in case of a redirect, navigationStart points to the start
> of the redirect (Point A), while the start of the
>     destination page is denoted by fetchStart (Point B).
>     - when navigate within the same origin, navigationStart is the starting
> point of the user experience.
>     - navigationStart will be protected by the same origin policy and is
> not always available.
>
>    Similar situation goes for unloadEventStart/End as well, i.e., should
> they refers to the unloading of the page prior to
> the redirection, or that of the page prior to the destination page (which
> might not need to be unloaded at all).
>
>    The current draft use the first proposal while IE and Chrome follow the
> second.
>

> cheers,
> Zhiheng
>