- From: Ricardo Oliveira <rvelosoo@gmail.com>
- Date: Wed, 16 Nov 2011 01:58:26 -0800
- To: public-web-perf@w3.org
Hi Tony, Thanks for getting back. Due to the significant number of samples where this is happening, i doubt it's related to clock adjustments (e.g. ntp, leap second, skew adjust, etc). In any case, looking forward to see this fixed. Cheers, --Ricardo On Nov 16, 2011, at 1:49 AM, Tony Gentilcore wrote: > On Wed, Nov 16, 2011 at 4:13 AM, Ricardo Oliveira <rvelosoo@gmail.com> wrote: >> Hi there, >> >> If I define: >> >> fetch_time = responseEnd - fetchStart >> dns_time = domainLookupEnd - domainLookupStart >> connect_time = connectEnd - connectStart >> request_time = responseStart - requestStart >> response_time = responseEnd - responseStart >> >> And if i do: >> >> reconnect_time = fetch_time - dns_time - connect_time - request_time - response_time >> >> I get several negative values over different versions of Chrome/FF. The idea was for the reconnect_time to measure attempts of failed dns lookups, tcp connects. For the case where everything is in cache, the idea would be to measure the cache lookup time... >> >> Is there any other way to compute this? Is this suppose to be negative? > > Your interpretation of the spec is correct. I believe this WebKit bug > is causing the problem for Chrome (not sure about FF): > https://bugs.webkit.org/show_bug.cgi?id=58354 > > Hopefully that will be fixed soon. > > In the meantime, it would be really helpful if we could write a > conformance test for http://w3c-test.org/webperf/tests/ that would > fail when this bug is hit. > > -Tony > >> >> Cheers, >> >> --Ricardo >>
Received on Wednesday, 16 November 2011 09:59:18 UTC