--MimeMultipartBoundary
Content-Type: TEXT/PLAIN; charset=US-ASCII
On Mon, 25 May 1998, Stewart Forster wrote:
> > Hm, perhaps I've missed something, but if your decision to ignore
> > "keep-alive" is based on User-Agent header alone, then _all_ proxies in the
> > request processing chain will ignore "keep-alive" because User-Agent is
> > passed through by a proxy unchanged. You might distinguish between
> > Proxy-Connection and Connection headers somehow (assuming clients never send
> > Proxy-Connection), but I doubt it is a robust solution.
>
> In fact you are correct. I had assumed that Squid would try to set
> up a persistent connection with the next hop regardless of what the client
> dictated. ie. client <- non persistent-> squid <- persistent -> web server
> I'm not very clued up on persistent HTTP in Squid but that was how I assumed
> Squid would have operated rather than letting the client dictating to the
> entire connection chain. Of course what I'm suggesting may make no sense
> or go against the HTTP 1.1 specification. Is that so? (I know - RTFM) :/
Hi Stewart,
I was just trying to say that your patch (and the first version of my
patch) may prevent Squid from establishing persistent connections between
_proxies_ if the browser originating the connection was Netscape 3.x.
You see, the _same_ code gets executed for every incoming connection on every
proxy (leaf, intermediate, or root), and User-Agent header stays the same all
the way through. In other words, the old code does not distinguish between
connections from a proxy and connections from a browser. Thus, we should take
special care _not_ to disable persistent connections based on User-Agent when
connection is not comming from a browser (i.e., another _proxy_ talks to us).
Current code attempts to do that, but I am not sure we covered all the bases.
Alex.
--MimeMultipartBoundary--
Received on Tue Jul 29 2003 - 13:15:50 MDT
This archive was generated by hypermail pre-2.1.9 : Tue Dec 09 2003 - 16:11:47 MST