Re: [NavigationTiming2] How many 'navigation' entries are available?

On Thu, Nov 27, 2014 at 11:45 AM, Jonathan Watt <jwatt@jwatt.org> wrote:

> The draft describes the creation of an PerformanceNavigationTiming
> instance in section 5, but it doesn't seem to say anything about how that
> object becomes available to scripts. It also doesn't say much about the
> sequence returned by performance.getEntriesByType("navigation") and what
> entries are added to it and when they become available.
>
> * Are entries from anything other than the last navigation available?
>

No, only the current/last navigation is available via performance.timing.


> * If so, what are the origin restrictions? If a non-same origin navigation
>   happens between two same origin navigations, does the sequence just not
>   contain the non-same origin navigations, or does everything prior to the
>   recent series of same origin navigations not appear in the sequence?
> * Do new navigations appear at the beginning or end of the sequence?
> * Does a back/forward destroy previous entries in the series, or just
>   add more to it?
>

None of these apply.


> I'm guessing browser vendors don't want to use memory keeping previous
> navigation entries around for the rare case that they might be used, so
> maybe the sequence always only consists of a single entry. If so the spec
> should say so explicitly though, and if not then the above should be
> clarified with normative text even if the desired behavior might seem
> obvious.
>

A quick spot check in Chrome and FF shows that both return an empty array
for window.performance.getEntriesByType("navigation"). Not sure what we
actually want here though... Perhaps for consistency with other events we
should return an array with a single entry? I could be convinced either way.

ig