Re: HTTbis spec size, was: Rechartering HTTPbis

lör 2012-01-28 klockan 11:50 +0100 skrev Julian Reschke:

> I *agree* with improving the layering, and maybe giving the transport 
> layer a specific name, but "HTTP/2.0" it can't be.

My vision of HTTP/2.0 is

- HTTP message definition, defined largely by what we have today minus
anything transport related, and trowing out any special cases. No
definition of message delimiting, no chunking, no connection management,
no folding, etc. That's all transport defined.

- Definition of required transport properties/interface for HTTP message
exchanges.

- Definition of suitable transports ontop of TCP, TLS, UDP, DTLS, SCTP.
Each separate, mapping the requirements of the above. Completely
ignorant of any semanitcs of HTTP beyond the most basic structure.

- Mapping guidelines how to map relations bewteen transport identity
assosciations and messages, i.e. how transport level authentication fits
with messages.

- Defined upgrade paths from HTTP/1 to HTTP/2. I.e. HTTP Upgrade, TLS
negotiation, SRV records, etc...


I do not see a great need for a new semantic model. But several things
that need to change on the wire to provide a robust & efficient protocol
that can be built upon for the next several decades.


Regards
Henrik

Received on Monday, 6 February 2012 03:48:36 UTC