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

On 2014-07-17 14:44, Poul-Henning Kamp wrote:
> In message <53C7C3CC.8080105@gmx.de>, Julian Reschke writes:
>
>> FWIW, I do support that we don't need 101, and that we might be able to
>> do 100 better.
>>
>> However, that doesn't cover other cases of 1xx. Yes, I realize that
>> there currently do not seem any, but I'm very unhappy to close an
>> existing extension point without very good reasons.
>
> I understand your unhappiness, but invoke Gettys rules 1 & 3:
>
> 1. Do not add new functionality unless an implementor cannot complete
>     a real application without it.

We're not talking about new functionality.

> 3. The only thing worse than generalizing from one example is
>     generalizing from no examples at all.
>
> I'll be happy to look at any concrete proposals if and when they
> appear, but anything we define in a vacuum would be widely unimplemented
> and therefore probably not work at all.

The concrete proposal is to handle 1xx messages in HTTP/2 exactly the 
same way as in HTTP/1.1. HTTP/2 will need code (and specification) to 
deal with 1xx one way or another, and I believe my proposal has the 
property of being both simple and consistent with 1.1.

Best regards, Julian

Received on Thursday, 17 July 2014 12:52:33 UTC