W3C home > Mailing lists > Public > public-web-perf@w3.org > November 2010

Re: [NavigationTiming] Move window.performance to navigator.performance

From: Jonas Sicking <jonas@sicking.cc>
Date: Tue, 23 Nov 2010 10:21:45 -0800
Message-ID: <AANLkTikm-fosrSVBkxfJuv1LAw-Dme3sTfW9hWExnJE4@mail.gmail.com>
To: Simon Pieters <simonp@opera.com>
Cc: "public-web-perf@w3.org" <public-web-perf@w3.org>
On Tue, Nov 23, 2010 at 7:17 AM, Simon Pieters <simonp@opera.com> wrote:
> I'll assume that you intended to cc the list...
>
> On Tue, 23 Nov 2010 11:59:55 +0100, Jonas Sicking <jonas@sicking.cc> wrote:
>
>> On Tue, Nov 23, 2010 at 2:05 AM, Simon Pieters <simonp@opera.com> wrote:
>>>
>>> Hi,
>>>
>>> I'm a bit concerned about
>>>
>>> http://test.w3.org/webperf/specs/NavigationTiming/#nt-window-performance
>>> [Supplemental]
>>> interface Window {
>>>  readonly attribute Performance performance;
>>> };
>>>
>>> This adds a new property 'performance' to the global scope. I think there
>>> may be pages out there that already use that property for other things --
>>> script variables or <div id="performance"> or <a name="performance"> or
>>> <form onsubmit="return validate(performance.value)"><input
>>> name="performance">, and so forth.
>>>
>>>
>>> http://google.com/codesearch?q=name%3D%28%22%7C%27%29%3Fperformance%28%22%7C%27%7C%5Cs%7C%3E%29+lang%3Ahtml
>>>
>>> To avoid polluting the global scope with a common name, potentially
>>> breaking
>>> scripts out there, I suggest we move the attribute to the Navigator
>>> interface.
>>>
>>>
>>> http://dev.w3.org/html5/spec/timers.html#system-state-and-capabilities:-the-navigator-object
>>>
>>> To follow the convention around Navigator in HTML5 I guess it could look
>>> something like:
>>>
>>> [Supplemental, NoInterfaceObject]
>>> interface NavigatorPerformance {
>>>  readonly attribute Performance performance;
>>> };
>>> Navigator implements NavigatorPerformance;
>>
>> One problem with using navigator is that the navigator instance
>> survives across same-origin page transitions. (At least in gecko)
>
> Interesting, I didn't know that. It doesn't seem to survive page transitions
> in Opera or WebKit, but in IE it does.
>
> Why does it survive page transitions in Gecko?

This is behavior that we carried over from Netscape 4. I suspect that
it's required for web compat. It allows pages to set "expandos" in one
page and see them on the next. I.e. a primitive sessionStore.

> Why is it a problem for navigator.performance?

It just makes behavior confusing since it means that the same
navigator instance will be morphing to reflect the various pages
performance info as you go back and forward through history.

> Is it also a problem for any of the other members of navigator?

None of the other properties on navigator are page specific, but
rather reflect properties of the browser. Which makes sense given its
name.

/ Jonas
Received on Tuesday, 23 November 2010 18:22:44 GMT

This archive was generated by hypermail 2.2.0+W3C-0.50 : Tuesday, 21 December 2010 18:13:56 GMT