Re: Our ALPN protocol IDs

On 2013-12-06 07:00, Martin Thomson wrote:
> On 5 December 2013 09:49, Carsten Bormann wrote:
>> Well, is HTTP/2.1 going to be “incompatible” with HTTP/2.0?
> 
> That's not a question that can be answered without a crystal ball, so
> how about I answer a different question.  If HTTP/2.1 turns out to be
> incompatible, it will need to be negotiated using a new identifier.
> If it turns out to be compatible, then we'll work something out.
> 
> I know that sounds a little unsatisfying, but it's the incompatible
> changes that hurt, and we have discussed a way to avoid bad stuff from
> happening when we need to do that.  Plotting out a path for compatible
> changes doesn't really add value.  That said, feel free to propose
> something.

If 2.1 turns out to be incompatible at the binary level from 2.0 it 
qualifies for being 3.0 in accordance with the versioning RFC.

Keep in mind that "compatible" with 2.x means only the frame header 
binary pattern (frame type, length, stream ID, flags) and semantics of 
handling unknown frames in 2.0.

So 2.1 may use completely different frame types and payload formatting 
and still be "compatible" enough to qualify for only a minor version 
bump.

Amos

Received on Thursday, 5 December 2013 22:35:28 UTC