Re: WCCP - cache only , no proxy

From: Lincoln Dale <ltd@dont-contact.us>
Date: Tue, 18 Apr 2000 05:13:07 -0400

At 01:26 18/04/00, Henrik Nordstrom wrote:
> > it's cache but upon retrieving the object , squid uses the client's IP
> > address who requested the object ...
>
>The hardest part is to get the return packets to go to Squid rather than
>the client IP. WCCP can redirect the client requests to Squid, but how
>do you redirect the packets arriving from the internet to Squid if Squid
>is using the client's IP address?
>
>I only see this possible when Squid is running on the border gateway or
>behind a smart L4 switch who tracks TCP sessions rather than IP/mac
>addresses, not when using WCCP.

in WCCP's case, it is a deliberate decision to NOT masquerade as the
web-browser to the web-server, but only in the direction of web-browser to
web-server.

the web-server will only ever see requests from cache's ip address itself,
and only ever from the real web-browser if the flow is bypassing a cache in
some manner (not really relevant to WCCPv1 but more-so for WCCPv2).

while it wouldn't be impossible to do this, and take the hit of maintaining
all that per-flow state information in a switching device (can you say:
"doesn't scale"?), it would break spectacularly if you had assymetric
routing or redundant paths that traffic can take back to the end-user.

cheers,

lincoln.

--
   Lincoln Dale           Content Switching
   ltd@cisco.com          Cisco Systems Inc.        |         |
                                                    ||        ||
   +1 (408) 525-1274      bldg G, 170 West Tasman  ||||      ||||
   +61 (3) 9659-4294 <<   San Jose CA 95134    ..:||||||:..:||||||:.. 
Received on Tue Apr 18 2000 - 06:19:08 MDT

This archive was generated by hypermail pre-2.1.9 : Tue Dec 09 2003 - 16:52:58 MST