We have a simple web application deployed on WebSphere 6.1.0.23 (Unix) - the app is secured with Spring 2.0.x security module and the back end is Spring 2.5.3 (just simple MVC).
We've developed a Custom PreAuthentication filter that validates user credentials and performs proper login. During integration testing we discovered that there were intermittent problems with authentication and that sometimes we were mysteriously re-directed to the Login page.
These problems were isolated to only IE (we're using 6.0.2900.x). Turns out that when you explicitly add the external domain name of our site to the site specific actions in IE (Tools > Options > Privacy > Sites > ALLOW our domain), then the web application's security functions perfectly. If the site is not explicitly in the list the security sometimes does NOT work in IE but it always seems to work in FireFox. This is obviously a problem for our users that we'd like to resolve on our end...
We've used the TCP/IP monitor to examine the headers that the browser is issuing in each case and although I can see some small differences between FF and IE but I do not see ANY differences in headers sent by IE when the Privacy settings are in place or when they are NOT in place - these two requests look identical in TCPIP monitor, but the results in our app are vastly different... What we can't figure out is what is different between those requests that our application 'sees' - we can see in the server logs that the request is validated so the login screen is a big mystery to us...
If someone could give some clues in this area I'd appreciate. See attached file for our security configuration...



