Re: HTTP/1.1 (was Re: NANOG outcome?)

map@iphil.net writes:

>Hi,
>
>Duane Wessels wrote in ircache:
>> What I got from the W3C's notes on pipelining is that HTTP/1.1
>> pipelining gets you a little better latency (20%?) but the real win
>> is that you transmit a lot less packets overall.
>
>Man, doesn't that make pipelining/persistency more attractive! :)
>
>The way it looks so far - correct me if I'm wrong - is that the
>current squid state machine architecture is geared towards the
>start/stop paradigm. Would it take an overhaul to support opening
>the TCP connections to the peers on startup, and attempting to reopen them
>when these fail?

I think its possible. It will just mean keeping additional
state information around, for each connection.

Duane W.
Received on Tue Jul 29 2003 - 13:15:40 MDT