- From: Ilya Grigorik <igrigorik@google.com>
- Date: Fri, 27 Feb 2015 11:06:15 -0800
- To: Todd Reifsteck <toddreif@microsoft.com>
- Cc: "public-web-perf@w3.org" <public-web-perf@w3.org>
Received on Friday, 27 February 2015 19:07:23 UTC
On Fri, Feb 27, 2015 at 10:38 AM, Todd Reifsteck <toddreif@microsoft.com> wrote: > Thanks, Ilya. The web interface is a bit tricky to navigate and I missed > that response. > > > > To summarize and ensure I’m understanding the argument: > > If 3rd parties implemented > http://www.w3.org/TR/resource-timing/#timing-allow-origin, > networkDuration is not necessary because sufficient information would > already be included. > > > > Is that accurate? > - For same origin resources you already have all the detailed metrics to compute networkDuration. - For first party origins you should enable TAO to get detailed metrics and compute networkDuration from them. - For third party origins that you don't control (correctly) exposing networkDuration circumvents TAO [1]. - networkDuration is a noop for HTTP/2. More generally, I think we should stay away from defining computed metrics on the perf interface. Our job is to provide the raw timestamps, and we should leave the rest to the application... networkDuration is just one of many possible computed metrics [2] that we *could* expose, so it quickly becomes a slippery slope of "why not also add X, Y, Z". [1] https://lists.w3.org/Archives/Public/public-web-perf/2015Jan/0002.html [2] https://lists.w3.org/Archives/Public/public-web-perf/2015Jan/0031.html
Received on Friday, 27 February 2015 19:07:23 UTC