I think that out of all the posts, mine ususally take the longest to get responses (If there are any). We'll probably going to continue the tradition with this one!

I have created a site that will function in SSL or NON-SSL mode. For some reason the script will never logout of the system (using the standard "do=log_off"). I have not changed any of the original coding in the original CGI's, everything is all template based.

I have created an account that (I will diasble later) will allow you to see what I am doing. It is weird because on some browsers the logout will work in non-SSL connection (on the site just delete the "s" in https) and some it won't. But no "logoff" function seems to work within the default SSL version of the site.

The site is a health services company and the DbManSQL templates have been "heavily" modified to fit our purpose. Also the site is a little slow just because of the SQL transmission overhead which will be resolved before the site goes live.

Site address: https://www.alliedhealthsystems.com/code/db.cgi

Username: 111111 and Password: admin

Please let me know what you may find out or if you have any questions with the site. Thank you!

Hi,

it might be too simple as approach, but:

1. I have tested what you said in both SSL/non-SSL version of the site and logging off didn't work with the newest Internet Explorer for both versions. I guess it has nothing to do with your SSL setup.

2. Have you tried to run the system without sessions (with cookies)? Can you log off properly then?

3. Since you're working with sessions, the system has to modify the user_sessions table in the db via SQL each time someone is logging in/off. You've mentioned that the db is slow - do you have a possibility to solve this speed issue before solving the log off problem?

Hope this helps,
Oliver

Well it's a great thought but I switched over to cookie based authentication and had the same problem with the logging out (the server just times out) in both SSL and non-SSL. Any further ideas?

Note: Switching to cookies actually broke several pieces of code I have because they rely upon the SID variable to determine where the script is. But by manually doing the "do=log_off" command I discovered the same scenario.

Any ideas, Stone/others?

Speed issue.

The only reason that the database is running a bit slow is because I'm running all the data across an ODBC bridge to a windows machine. The SQL/windows machine is located across a frame connection so the latency is somewhat high (about 11ms). I'm working on the speed problem (by moving the webserver behind the corporate firewall but I'm a week or two away from that) but highly doubt that is the problem as ALL other functions work just fine (including add, search, delete functions). Even add field functions from the admin interface.

I don't know how the log off thingy should be done in a correct manner, but after doing some tests, I could sucessfully log off for the first time using the following URL:

https://www.alliedhealthsystems.com/...1&Password=admin

check it out, Oliver

Yes, that does work for me as well... except it does not take the session data out of the SQL table. All of the session information stays. If I add the SID info we go back to where we started. This proves that the log_off coding works it is just something weird with the SQL portion of the log_off coding???

I do know that out of experienence it will NOT logoff unless I have a Username and Password associated with the path info. I have tried many different variations (with/without SSL, with/without username & pass, etc.) And I cannot resolve the problem. I think I am going to send in a support ticket on this because it is appearing more and more to be a bug of some type.

hmmmm... I've noticed that you're having links to several (different) tables on the main screen. I don't know if the problem has something to do with this. The correct syntax for session links (you're using it now) is sid=xxxxxxxxxxxxxxxxxxxxxxxx&do=whatever&db=tablename -> it's working without any problem on my server (I'm using the newest DBMan with ONE table). In your original post, you said that you've just modified the templates; I assume the code of the Perl part of DBMan hasn't been modified. Have you ever tried to use the default templates that are coming with DBMan? Can you log off using them?

Keep me updated once you know more...

Good luck,
Oliver

I'm just closing out this thread here as I seem to have fixed the problem.

It seems that after moving to a brand new server with the latest PERL DBI code everything works just fine now. I think it has something to do with the PERL DBI that was causing the problem because I moved the site over exactly as it was on the other server and it just started working (logging off that is). When in doubt... reinstall.