RE: [squid-users] Re: Implementation issues

Here at my work all employees sign a form stating that they realize email,
internet usage, file storage systems, etc are for work purposes and that
none of it is truly private. That email will not be directly read except for
troubleshooting/maintenance issues unless there are illegal or non-condoned
company activities taking place.

All traffic encrypted? Encryption puts a burden on the servers and if all
the traffic is being encrypted, then what exactly *is* being monitored?

Thoughts? Set your PC's gateway to your Apache server, visit www.google.com,
get your disclaimer page to come up, sign in, and then have it make you end
up at google. If this works, insert Squid afterwards. (one piece at a time,
see what works, what doesn't)

Chris Perreault

-----Original Message-----
From: Rick Whitley [mailto:rickwh@dbu.edu]
Sent: Wednesday, July 21, 2004 5:37 PM
To: squid-users@squid-cache.org; Chris Perreault
Subject: RE: [squid-users] Re: Implementation issues

The disclaimer page gives us the opportunity to inform the users that their
traffic is being monitored. It's the law. I am open to suggestions as to a
better way to accomplish this. I agree its ugly. The disclaimer page would
have a login link that went to the 2nd proxy. On top of everything, they
want the traffic encrypted. I need to get a workable process working first.
What if the first stop was a web server?

client -> Apache -> Proxy -> internet
                   |
               signon

Here apache would trap the traffic unless authenticated(session variable?).
We would have apache be the gateway.

Thoughts?

rick...
Rom.5:8

>>> Chris Perreault <Chris.Perreault@Wiremold.com> 7/21/2004 3:55:21 PM
>>>
First attempt:

client -> Proxy1 -> sign-on/disclaimer -> Proxy2 -> ldap -> Internet

User clicks on a page, sending another request, which then will do
this:

client -> Proxy1 -> sign-on/disclaimer -> Proxy2 -> ldap -> Internet

If proxy1 sends all traffic to the signon page, then all traffic passes to
it. All of it. Each request is coming from the browser, not the signon
server, so each request hits the proxy1 first and gets resent to the logon
page. Ok...so if a session variable knows the user logged in, then the
signon server can redirect the request for the website through proxy2. All
requests would pass through proxy1-->signon server-->proxy2. If that's
acceptable, it looks like it would work. You'd want to redo the session
variable on each hit, or else you could have the user relogging in every 20
minutes or so. If they were doing anything on the web (filling in a form or
taking a test) and didn't submit another request for a while, their session
could time out and they sure wouldn't be happy. It looks doable, but it
looks ugly too.

Once a user has signed up, you want them to get the disclaimer page every
time they fire up a web browser?

It would be much easier to have them just sign an "internet use policy" and
in exchange tell them their username and password:)

Chris Perreault
Webmaster/MCSE
The Wiremold Company
West Hartford, CT 06010
860-233-6251 ext 3426

-----Original Message-----
From: Rick Whitley [mailto:rickwh@dbu.edu]
Sent: Wednesday, July 21, 2004 3:44 PM
To: squid-users@squid-cache.org; Chris Perreault
Subject: RE: [squid-users] Re: Implementation issues

What we are thinking of doing is:

client -> Proxy -> sign-on/disclaimer -> Proxy -> ldap -> Internet

The 1st proxy will be open and require no auth and redirect all traffic to
the sign-on/disclaim site. User has option to Activate account or visit
Internet. The "Visit Internet" link will change go to the 2nd proxy which
will have proxy_auth enabled for ldap. The client will be prompted with a
userid and passwd dialog to be authenticated and sent to the internet.

Does this seem possible? I have the gateway for the segment set up to be the
1st proxy so I may still have a loop issue. Is there an automated way to
modify the gateway on client systems? Like a forced dhcp without the
request. If we could make this behave as two separate segments ( seg1 = 1st
proxy with no auth required, seg2 = 2nd proxy with auth
required.) client would always start in seg1 and have to request seg2. Any
thoughts or suggestions are greatly appreciated!!

thanks

rick...
Rom.5:8

>>> Chris Perreault <Chris.Perreault@Wiremold.com> 7/21/2004 2:02:23
PM
>>>
Doh, there is a referer_regex ACL type, but I don't see that helping here
anyways.

1)Browser wants to visit website.com and hits the proxy.
2)Request gets redirected to the signup_disclaimer.htm website. 3)User signs
up and/or logs in. 4)The page that verifies the username/password then
redirects to the originally requested site with the login information stored
in the correct header. (via the proxy)
5) the proxy sees the refering site and proxies the request.

A better 5) would be squid sees it has an already authenticated request so
passes it through. Otherwise the user would have to log in, on the
disclaimer page, every time they clicked a link. Open a new window/session
though and then the user would have to log in again.

Hopefully my ramblings help Rick out. I see a loop happening though. If
proxy_auth is required, the user gets the pop up window, even if they don't
have an account yet. Once they visit the login page, that page should be
able to write/create the http_proxy_authorization header.
I don't see how to redirect the user if they are not auth'd yet without
redirecting them when they are auth'd later so it looks like they are stuck
with the logon prompt and only get to the disclaimer/login page when they
fail to authenticate.

Chris Perreault

-----Original Message-----
From: Chris Perreault
Sent: Wednesday, July 21, 2004 10:15 AM
To: squid-users@squid-cache.org
Subject: RE: [squid-users] Re: Implementation issues

We may turn into such a sponser. In reverse proxy mode we are looking for a
solution where we need to authenticate users as far out in the DMZ as
possible. Using ldap via basic auth results in some difficulty in reaching
some of the content on various back end webservers which also uses basic
auth. Over the next few days we'll determine the best solution, which may be
having users redirected to a logon webpage, use a form based authentication
that saves the ip, then passes the username down in a header to the back end
webservers, thus allowing some of those servers to match the header username
with a username in a profile database for portal type content delivery while
still allowing basic auth type apps to do their own authentication.

If squid knew the referring IP address (the webserver that has the
authenticating form on it that after submital send the browser back to
Squid) then all users could be directed to the logon page, and after hitting
submit go back to the proxy which would allow that referring IP to get
proxied info from the web. Spoof the IP though and you get through without
authenticating. Figuring that IP out might be difficult for a end user
though.

Chris Perreault

-----Original Message-----
From: Adam Aube [mailto:aaube01@baker.edu]
Sent: Tuesday, July 20, 2004 8:15 PM
To: squid-users@squid-cache.org
Subject: [squid-users] Re: Implementation issues

Henrik Nordstrom wrote:

> On Wed, 14 Jul 2004, Rick Whitley wrote:
>
>> I was setting up a proxy server to do the authentication and
>> cacheing, but have learned from the list that it is not going to
>> behave the way I expected. Users should only see the initial page
>> once. I seem to be out in left field as to how to implement this.
Any
>> suggestions?
>
> In theory it is possible to implement very close to what you want
with
> Squid & authentication, but so far no one has been willing to
actively
> sponsor such development.
>
> It is also possible to use a small proxy.pac script implementing the

> "startup splash" screen.

Something like this might be a good starting point:

http://www.squid-cache.org/mail-archive/squid-users/200404/0793.html

Adam
Received on Thu Jul 22 2004 - 05:42:58 MDT