cgi pro*c and kill

Hello,

I've got a problem using Pro*C programs with the Netscape Enterprise 2.0 web server.

Scenario:

If a web browsing user presses the 'STOP' button on their browser during a db access, the cgi program hangs and holds onto all locks it has on the db. To get around this I have written a program that gets called each time a program connects to Oracle, and kills any processes
(unix) that may still be running as a result of the user pressing
'STOP' previously.

Problem:

The process that is the target of the kill (unix) does die but there are STILL locks on the Oracle db. A process called 'pmon' eventually notices (usually within a minute) and kills the shadow Oracle process
(I think) which clears all of the locks. The problem is the db is
still locked until pmon does its biz.

Help:

Does anyone know if this is how it must work; or can anyone think of a neater, working solution that will realise that the web browsing user has hit stop so that we can abort the backend Oracle process. It appears that NO signals are sent from the Netscape server to the cgi program when stop is pressed.

Many thanks,

Ian.

--
Ian Steel
ian_at_bilstone.co.uk