Re: END_SEGMENT and headers

The draft states both that frames can not be coalesced across a segment
boundary, but also that segment boundaries must be preserved.

You are correct that for a HEADERS frame w/ the END_SEGMENT flag set, there
is no way a server could coalesce frames because the next frames must be
CONTINUATION frames until the END_HEADERS flag is received, so in this
case, END_SEGMENT and END_HEADERS are equivalent.

BUT

For a frame with the END_SEGMENT flag CLEARED, the HEADERS/CONTINUATION
frame sequence would still need an END_HEADERS flag set, so in this case
they are not equivalent. In this case, combining the flags would be
inserting a segment boundary where none previously existed.


On Wed, Apr 16, 2014 at 4:36 PM, David Krauss <potswa@gmail.com> wrote:

>
> On 2014-04-17, at 7:33 AM, David Krauss <potswa@gmail.com> wrote:
>
> > a CONTINUATION frame must be preceded on the connection by a HEADERS or
> CONTINUATION frame from the same stream.
>
> To be clear: *immediately* preceded with no intervening frames. Also, a
> DATA frame is not valid until after END_HEADERS which moves to the "open"
> state.
>
>
>