Re: Framing and control-frame continuations

Agreed, and the trend is that is only increasing. Using a bit to indicate
end-of-stream (at least unidirectionally, assuming that can be done safely)
is most efficient in such cases.
Were it the other way around, though, of course, it might be more efficient
to do it another way.

-=R


On Fri, Feb 8, 2013 at 12:49 AM, Adrien W. de Croy <adrien@qbik.com> wrote:

>
>
> ------ Original Message ------
> From: "Roberto Peon" <grmocg@gmail.com>
> To: "Martin Nilsson" <nilsson@opera.com>
> Cc: "HTTP Working Group" <ietf-http-wg@w3.org>
> Sent: 8/02/2013 8:32:47 p.m.
> Subject: Re: Framing and control-frame continuations
>
>  Not in cases where one side of a flow often closes after one control
> frame, e.g. most HTTP GETs
> -=R
>
> IME most GETS are followed by another GET.  HTTP/1.1 anyway.
>
> Adrien
>
>
>
>
> On Thu, Feb 7, 2013 at 11:26 PM, Martin Nilsson <nilsson@opera.com> wrote:
>
>> On Wed, 06 Feb 2013 12:47:08 +0100, Amos Jeffries <squid3@treenet.co.nz>
>> wrote:
>>
>>
>>> The Frame format:
>>>
>>>          0                   1
>>>          0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5
>>>         +-+-+-----------+-------------**--+
>>>         |F|C|  type     |               |
>>>         +-+-+-----------+               +
>>>         |        Frame Length (24)      |
>>>         +-----------------------------**--+
>>>         |       opaque ID (16)          |
>>>         +-----------------------------**--+
>>>         |     Frame Data (16...N)       |
>>>         +-----------------------------**--+
>>>
>>>
>> Since the flow only ends once, isn't an end-of-flow control type more
>> efficient use of bits than a flag?
>>
>> /Martin Nilsson
>>
>>
>