> Ok... here's a status report. In trying to debug this thing, we
> disabled the "transparency" mode by changing the firewall configuration
> to NOT route all HTTP traffic to the squid-cache and instead run HTTP
> traffic through itself as it did before. I then reconfigured my browser
> settings to force me to go through the proxy-cache and tried reaching
> the same troubled web sites. THEY WORKED!! This leads me to believe
> that there is something wrong with the way I'm setting up the
> "transparency" configuration. The Watchguard firewall allows me to
> enter the IP and Port of a proxy server. While reading through the
> "transparency" how-to, they tell you to create a route on the
> proxy-server back to the firewall. But my firewall and proxy-server are
> on the same subnet. So by simply keeping the default route that is on
> the proxy-server, it finds the firewall anyways. But this is where I
> start to doubt if I'm correct.
>
> Do you have any ideas??
Not really, only the list of possible caveats which can be encountered
with transparant proxying, therefore I advice against it :
- Intercepting HTTP breaks TCP/IP standards because user agents
think they are talking directly to the origin server.
- It causes path-MTU to fail. Possibly making the website not accessible.
- As a result for instance on older IE versions ; "reload" did not
work as expected.
- You can't use proxy authentication
- You can't use IDENT lookups
- Intercepting proxies are incompatible with IP filtering designed
to prevent address spoofing.
- Clients are still expected to have full Internet DNS resolving
capabilities , when in certain Intranet/Firewalling setups , this
is not always wanted.
- Related to above : because of transp. proxy setup : squid
connects to a site
which is down.HOWEVER , due to the transparant proxying setup. It gets
a connected state to the interceptor. The
end user may get wrong error messages or a browser, seemingly
doing nothing anymore.
>
> The firewall and squid run on a 10.0.200.0/24 subnet (This is the subnet
> we keep all our main servers on)
> Workstations use a 10.0.204.0/22 subnet
> Render Farms and Storage use a 10.0.108/22
>
> Do I need to put a route into the squid for each of the other networks
> that point back to the firewall? (10.0.204.0/22, 10.0.108.0/22)
>
I don't think so.
M.
Received on Thu Jan 19 2006 - 09:53:53 MST
This archive was generated by hypermail pre-2.1.9 : Wed Feb 01 2006 - 12:00:01 MST