Re: agenda/charter brainstorming

In message <53A7E8BA.8090809@gmx.de>, Julian Reschke writes:
>Hi there,
>
>here are some ideas for what the WG could/should work on in the 
>post-HTTP2-LC time:

HTTP/3.0
--------

A non-backwards compatible, heavily cleaned up HTTP and much simpler
HTTP protocol for the longer future:

Goals:

    Scale towards Tbit/s speeds on contemporary silicon
	Fiber to the home is coming, we need to deal
	with it.

    Improve privacy in ways which does not force a crypto-arms-race.
	Indiscriminately adding crypto in the face of laws
	just means that the crypto will be circumvented or trojaned.
	We need to pay the king his shilling, in a way that means
	he doesn't have to take the entire pound.

    Reduce cruft and simplify the semantics of HTTP.
	HTTP has become too complex through accumulation of
	successive "convenient quick fixes" many of which
	nobody uses and most of which don't work reliably
	end-to-end any more.

Means:

    Define "routing-envelope" fields always sent in plaintext
	This enables load-balancing, also of otherwise secured
	transactions.  Eliminates need for certs on load-balancers

    Reduce the number of headers proxies need to pay attention to
	For speed and privacy.
	
    Replace cookies with client chosen session identifiers
	This increases transmission performance a LOT and
	solves the problem EU's "cookie directive" tried to
	address.

    Allow privacy protected and plaintext requests on the same connection.
	Less connection-churn when upgrading.
	Better connection-aggregation by proxies.

Consequences:

    Semantic changes will force website designers to think more
    clearly about privacy and thus this will not be a direct
    "invisible" upgrade from prior HTTP versions.
	

    

-- 
Poul-Henning Kamp       | UNIX since Zilog Zeus 3.20
phk@FreeBSD.ORG         | TCP/IP since RFC 956
FreeBSD committer       | BSD since 4.3-tahoe    
Never attribute to malice what can adequately be explained by incompetence.

Received on Monday, 23 June 2014 09:06:46 UTC