Re: Getting to Consensus on 1xx Status Codes (#535)

On 18/07/2014 1:36 a.m., Poul-Henning Kamp wrote:
> In message Michael Sweet writes:
> 
>> I'd just call the flag "EXPECT_WINDOW_UPDATE" and document that it
>> corresponds to the Expect: 100-continue semantics of HTTP/1.1.
> 
> Works for me.
> 
>> We'll still need to define the HTTP/1.1 gateway behavior, e.g., a 2.0 to
>> 1.1 gateway needs to implement a timeout that sends a WINDOW_UPDATE
>> frame if it doesn't see a 100 response in a timely manner.
> 
> The HTTP/1.1 spec already tells you that.
> 

Maybe I missed it. Where does HTTP/1.1 spec mention WINDOW_UPDATEs exactly?

The HTTP/2 spec needs to say *how* the mechanism maps to
Expect:100-continue at both gateway directions or we will end up with
implementations which forget or omit one half the mechanism. Thus
repeating the 100-continue problem all over again.

Amos

Received on Thursday, 17 July 2014 15:39:16 UTC