On 27 Nov 2003 07:57:45 -0800, [EMAIL PROTECTED] (Thelmo Loisio) wrote:

>On Wed, 2003-11-26 at 17:03, Greg Hennessy wrote:
>
>> You cant use a transparent proxy with a pure bridge. The connection to
>> remote site goes out from the proxy server not the the client.
>
>Well, if you assign to the nics and IP address you can, but maybe this
>for you isn't "pure bridge".

Thats it, you're now at layer 3 rather than Layer 2. 


>I've made another test, squid listen only on 127.0.0.1:3128 

Thats the way I have it here. If you're not seeing anything in the squid
logs, it sounds like the packet filtering is not quite right.  


KSF="keep state flags S/SA"
TCP="inet proto tcp"
UDP="inet proto udp"

# Add redirect to allow transparent caching of port 80 traffic.
#
rdr on $Inside proto tcp from $Lan to !$Lan port www -> 127.0.0.1 port 3128


Then the following. 

# Localhost
#
pass quick on lo0 $TCP all $KSF
pass quick on lo0 all keep state

Then 

# allow but dont log the following
#
pass out quick on $Outside $TCP from ($Outside) to !$LAN port http user \
_squid $KSF queue (q_def, q_pri) label "ACCEPT: proxy 


I havent used PF @ L2, I prefer working @ L3. 

>the client
>is configured to have the proxy on one ip assigned to one nic on the
>bridged box and on the bridged box i've a rdr rule that catch the
>connection to that ip... well in this configuration everything is
>working good.

Thats inline rather than transparent. 

>
>Any hint is really appreciated. Thanks

two questions, 

you've compiled squid with --enable-pf-transparent enabled ? 

Have you followed the instructions here.

http://www.benzedrine.cx/transquid.html

If you just want to implement a transparent cache, putting a route map on
your inside cisco and use policy based routing to hand off to squid would
be the way I'd do it. 


greg




-- 
$ReplyAddress = Use google to figure it out. 
The Following is a true story.....
Only the names have been changed to protect the guilty.

Reply via email to