W3C home > Mailing lists > Public > public-web-perf@w3.org > April 2011

RE: [agenda] Web Performance WG Teleconference #28 Agenda 2011-04-13

From: Nic Jansma <Nic.Jansma@microsoft.com>
Date: Wed, 20 Apr 2011 17:13:16 +0000
To: Tony Gentilcore <tonyg@google.com>, Zhiheng Wang <zhihengw@google.com>
CC: "public-web-perf@w3.org" <public-web-perf@w3.org>
Message-ID: <F677C405AAD11B45963EEAE5202813BD19D13A7F@TK5EX14MBXW651.wingroup.windeploy.ntdev.microsoft.com>
Look great.  Thanks Tony.

The only change I would propose is a change we just now submitted for all of the other tests -- bailing out early (and ensuring a single primary Fail) for UAs that don't have window.performance defined.

For example:
                test_namespace('navigation');
                if (performanceNamespace === undefined)
                {
                    // avoid script errors
                    done();
                    return;
                }

I'm in agreement with you regarding navigationStart/fetchStart -- let's discuss on the conf call today.

- Nic

-----Original Message-----
From: Tony Gentilcore [mailto:tonyg@google.com] 
Sent: Wednesday, April 20, 2011 9:51 AM
To: Zhiheng Wang
Cc: Nic Jansma; public-web-perf@w3.org
Subject: Re: [agenda] Web Performance WG Teleconference #28 Agenda 2011-04-13

On Wed, Apr 13, 2011 at 7:08 PM, Zhiheng Wang <zhihengw@google.com> wrote:
> * If there is no previous document, navigationStart must return the same value as fetchStart. unloadEventStart/End must return 0.

As promised on last week's call, I submitted a test for this. Please review:
http://w3c-test.org/webperf/tests/submission/Google/NavigationTiming/test_no_previous_document.htm

It only checks that unload times are 0, not that navigationStart == fetchStart. I'd like to raise a issue/question about that.

In section 4.2, the spec says "If there is no previous document, this attribute must return the same value as fetchStart." However, the more descriptive processing model (section 5.1) describes them being recorded as two distinct times. That sentence might typically be true when there are no 3xx redirects, but certainly isn't true for redirects. Even in the non-redirect case, if two distinct times are recorded back-to-back, there is a non-zero chance that the second one will be later than the first. I propose we stick with the processing model and drop that sentence from 4.2.

-Tony
Received on Wednesday, 20 April 2011 17:13:44 UTC

This archive was generated by hypermail 2.3.1 : Tuesday, 6 January 2015 21:04:30 UTC