On 2014–07–21, at 4:04 PM, Poul-Henning Kamp <> wrote:

> I think it makes everything much easier if END_STREAM always sits
> on the packet which closes the stream.

HTTP frames aren’t packets. The implementation needs a persistent internal flag to signal cleanup at the end of the block, whether the END_STREAM is received at the start of a frame, jumbo frame, super-multi-frame, or whatever.

Since END_STREAM is only defined for HEADERS and DATA, I don’t think it’s fair to say that headers make a special case.

CONTINUATION is a special case in other ways, of course, and I hope we can see through superficial differences and give jumbo HEADERS frames a size limit pass, rather than attempt the same with a convoluted definition of what a “frame” is.

Received on Monday, 21 July 2014 08:46:19 UTC