Re: Re: problems with the squid-2.5 conn

From: Henrik Nordstrom <henrik@dont-contact.us>
Date: Thu, 20 Apr 2006 15:01:49 +0200

tor 2006-04-20 klockan 19:11 +0800 skrev Steven Wilton:

> Do browsers send requests to different servers down the same TCP connection
> when proxies are set?

Yes, and this is an area open for exploration on what IE does when
connection-oriented auth is combined with proxies...

but I suppose that with your approach it will be happy no matter what.

> If browsers do exhibit this behaviour, my patch would cause all subsequent
> requests on the same TCP session to have the auth, pinned and must_keepalive
> flags set.

Yes. Not much to be done about unfortunately. The main downside is that
there might be a few requests not getting cached which otherwise would
have been if the client isn't behaving and opening a new connection when
visiting another site after using connection-oriented auth.

A bigger problem that needs addressing is that you should somehow make
sure to close the server-side connection when the client connection is
gone. This to avoid excessive number of tied up server-side connections
which can not be referenced any longer..

> I'm also wondering if your connection pinning work addresses this issue.

My approach had a very close relation between client<->server
connections, and barfed on the client if it did not behave. This is the
area where Adrian run into stability issues..

Regards
Henrik

Received on Thu Apr 20 2006 - 07:02:02 MDT

This archive was generated by hypermail pre-2.1.9 : Mon May 01 2006 - 12:00:03 MDT