It could be any of the ideas posted here, but they'd have to be implementation bugs. It'd be nice to be able to pin this down to specific UAs and pages, so we can debug them. On Thu, Sep 6, 2012 at 5:43 AM, Bryan McQuade <bmcquade@google.com> wrote: > Could it be that the DOM Interactive is triggered when closing </html> > is parsed but that for some pages: > 1. there are additional bytes in the HTTP response after the closing > </html> > 2. the time it takes to reach the last byte on the network is after > the time the closing </html> is parsed by the renderer? > > I have not read the spec in detail so this may not actually be a > possibility but I wanted to put it out there. > The parser is supposed to run until EOF. If there's anything after </html>, it'll just treat it as if it were inside <body> again. > I do also wonder how last response byte timing is handled for > responses that don't have a content-length or use chunked encoding and > for whom the connection is not immediately closed when the last byte > of HTML is received. > The recording of the last byte should take place in the network layer, which would follow the HTTP rules and would cover these cases. James
Received on Thursday, 6 September 2012 18:39:41 UTC