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:[EMAIL PROTECTED] Sent: Wednesday, July 21, 2004 5:37 PM To: [EMAIL PROTECTED]; 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 <[EMAIL PROTECTED]> 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:[EMAIL PROTECTED] Sent: Wednesday, July 21, 2004 3:44 PM To: [EMAIL PROTECTED]; 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 <[EMAIL PROTECTED]> 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: [EMAIL PROTECTED] 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:[EMAIL PROTECTED] Sent: Tuesday, July 20, 2004 8:15 PM To: [EMAIL PROTECTED] 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