- From: Arvind Jain <arvind@google.com>
- Date: Sun, 25 May 2014 20:00:03 -0700
- To: Anne van Kesteren <annevk@annevk.nl>
- Cc: Jonas Sicking <jonas@sicking.cc>, Jatinder Mann <jmann@microsoft.com>, "public-web-perf@w3.org" <public-web-perf@w3.org>, Mike West <mkwst@google.com>
- Message-ID: <CAOYaDdPuG4sJMcKEz0vdc3PBjOF1pZQ8VEAa06PE0OAVaDgbYg@mail.gmail.com>
OK I'll make the change for the language as you suggest. Re. 407, what is your suggestion. I'm not sure what should be done. On Fri, May 23, 2014 at 10:07 AM, Anne van Kesteren <annevk@annevk.nl>wrote: > On Fri, May 23, 2014 at 7:43 AM, Arvind Jain <arvind@google.com> wrote: > > On Thu, May 22, 2014 at 10:34 AM, Anne van Kesteren <annevk@annevk.nl> > > wrote: > >> Can you at least remove the bit about headers and redirects? There's > >> no platform API that can configure not to follow redirects so there > >> should not be any confusion there. > > > > I understand that leaving out this sentence keeps the spec fully > specified. > > But in a normal case of fetching a request, you'd use the entity body of > the > > response. In this case, we want to discard it - to the extent that the > > browser could even close the connection after receiving the header. When > we > > started discussing the spec, there were many questions explicitly about > what > > would the browser do with the entity body, and whether a 302 would be > > followed for the Beacon request or is it strictly a fire and forget > model. > > Sure, but in the end you are using fetch and all this follows from > that algorithm. This should be no more than an informative note, > definitely not a requirement. I could maybe see a case for something > along the lines of "Once a response is a returned, the fetch MAY be > terminated.". (You don't get a response until redirects are followed > and given you follow redirects you need to get at least that far.) > > > >> When there's a 407 response, whether there's going to be an > >> authentication dialog and whether we want to avoid that in this case > >> because the site might already have gone away. In general we don't > >> really have a good story for this. Mostly curious how sendBeacon() > >> implementations are handling this. > > > > We definitely don't want to show the authentication dialog just like we > > don't want to do it for 401. > > Are you sure? As this actually indicates a problem in the user's > setup. Would also be the only API that does this as far as I know. > > > -- > http://annevankesteren.nl/ >
Received on Monday, 26 May 2014 03:00:31 UTC