Re: multiplexing -- don't do it

In message <A4080B50-60F0-4EF5-A437-416EB4E72232@gbiv.com>, "Roy T. Fielding" w
rites:

>Regardless, I consider some form of multiplexing to be a requirement for
>whatever replaces HTTP/1.1, [...]

Seconded, but for different reasons.

First of all, I think we'll see more proxies than currently, because
Ipv4/IPv6 and HTTP-1.0/HTTP-2.0 proxies will be deployed.  Proxies
are in a much better shape to exploit mux than clients are.

Second because it is a much better and more general mechanism than
pipe-lining, while having the exact same properties with respect
to queuing etc.  (I'm prefectly happy with the default level of
mux available is one, until the destination indicates otherwise
by some means in the protocol.)

Third, because mux is the only trick in the book I can see which
will allow us to transfer both TLS and non-TLS in the same TCP
connection.  This is important both for proxies and for sesions
being upgraded from plain to TLS.

Finally because it gives us a growing room we need for the wacko
ideas of the next twenty years.

-- 
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 Tuesday, 3 April 2012 23:43:36 UTC