Adrian wrote:
On Mon, Mar 17, 2008 at 9:25 PM, Amos Jeffries <squid3@xxxxxxxxxxxxx> wrote:
Squid does not differentiate the types of auth a user has done.
It tries all methods its configured with (in the order configured) until
one succeeds. The common way to do this appears to be to use the
least-accepting method first and failover to the most-accepting. Or
vice-versa depending on the situation.
I want to put 'trusted' users through NTLM fakeauth so I can capture
their usernames without bothering them with a popup auth box. For the
'untrusted' user subnets, I want to give them a popup box and make
them authenticate.
Since fakeauth will always pass, I can't just configure the schemes in
succession. I was thinking of writing my own fakeauth code which
rejected anything in my 'untrusted' IP list forcing it to the next auth
scheme, but I don't think the IP address is passed to authenticate
scheme by squid to check against?
Any other ideas?
Thanks,
Adrian.
Three Squid instances. The front one (that everyone connects to)
decides which parent to use depending on src IP. Trusted users are sent
to an instance set up using fakeauth, others are sent to the third
instance which uses basic auth. The two parents don't have to do any
caching, or any logging (that can all be handled by the first instance),
but are just there to provide authentication details.
I give no guarantees of the stability or scalability of this set-up, but
it should do what you want.
Chris
