- From: Sigbjorn Finne <sof@opera.com>
- Date: Mon, 18 Aug 2014 09:22:21 +0200
- To: Arvind Jain <arvind@google.com>
- CC: public-web-perf <public-web-perf@w3.org>, Adam Barth <abarth@chromium.org>, David Benjamin <davidben@chromium.org>
Den 18.08.2014 09:05, skreiv Arvind Jain: > Got it. Given how we wrote it, it would be before the Fetch is > initiated. So it would be before the request is sent to the network > stack. > Thanks much for clarifying. For implementations that don't delay initiation (I know of none that do, but perhaps there are), this would amount to "Beacon-Age: 0" for all practical purposes. Which brings up the second question raised initially - what benefit does this bring? --sigbjorn > On Sun, Aug 17, 2014 at 11:39 PM, Sigbjorn Finne <sof@opera.com> wrote: >> Den 18.08.2014 08:14, skreiv Arvind Jain: >> >>> #11 is left to the user agent. In the usual case, it is hopefully soon >>> after step 10, but it can de delayed. Is it what you are asking? >>> >> >> It's in that area. So a UA is allowed to sample the time ("current time") at >> any point after it decides to go ahead with the Beacon request that >> sendBeacon() initiated, and there being no assumption that any/all >> client-side&measurable delays have been performed when it does so? >> >> --sigbjorn >> >> >>> On Sun, Aug 17, 2014 at 10:12 PM, Sigbjorn Finne <sof@opera.com> wrote: >>>> >>>> Den 18.08.2014 04:03, skreiv Arvind Jain: >>>> >>>>> requestTime should be set to the time immediately after sendBeacon() >>>>> is called. So it should be before the actual Fetch is performed. The >>>>> actual Fetch may be performed quite a bit later (e.g. if the user >>>>> agent decides to delay the transmission). >>>>> >>>> >>>> That makes sense, what about the other end of the interval that the age >>>> value captures, when is it assumed sampled, i.e., when is step 11 >>>> performed? >>>> (That was what I was referring to, sorry about confusing the spec >>>> naming.) >>>> >>>> --sigbjorn
Received on Monday, 18 August 2014 07:22:51 UTC