Re: Striving for Compromise (Consensus?)

On 12/07/2014 8:38 a.m., Poul-Henning Kamp wrote:
> In message Johnny Graetting er writes:
> 
>>> I mean that if we insist the entire header-set goes into a single
>>> frame at most *zero* set of headers can be incomplete at any point
>>> in time.
>>
>> False. Just because I declare a frame size doesn't mean I have to send all
>> of it.
> 
> If you ever want to use the connection for anything else, you'll have to.
> 

If you advertise a large size using Greg et al frames then don't send it
the recipient will 431 respond (if its over their limit), PROTOCOL_ERROR
on a timeout. No other party or connection was harmed in any way by your
action.


If you send a HEADERS without END_HEADERS flag in h2-13 then pause, you
stall the recipient and corrupt their HPACK state.
 Also, any recipient which performs fragment streaming (eg Roberto's
implementations apparently) you will stall and corrupt the HPACk state
of the recipients outbound connections. Thus stalling all other client
connections it has interleaved.

Amos

Received on Saturday, 12 July 2014 04:22:46 UTC