In article
<[log in to unmask]>,
[log in to unmask] says...
> Joe is correct. RPMCREATE does indeed create a session. This would not
> help getting around the user count issue. When I mentioned the use of
> sockets to PC clients, I left the security validation as a task for the
> reader. Actually I have done this whole process. A master process wait
> for connects to a well known (at least in our case) socket. When the
> connect completes, the open socket is handed off to a newly created
> process to service requests. At this point the only request that can be
> accepted is a pseudo-logon, in other words, it uses AIFCHANGELOGON with
> the passed logon string to validate a user and put the process in the
> appropriate account. After passing this logon, the user can to have all
> requests serviced. Note that AIFCHANGELOGON, just validates user/account
> and passwords and does not actually create a session. It does run the
> process as though you were logged on to that account and user.
>
> Hope this helps,
> John Zoltak
> North American Mfg Co
> Cleveland Ohio
>


John one problem I found in the past with AIFCHANGELOGON and
CREATEPROCESS is that you shared one CI session.   Therefore if you
built a temporary file, for example, then someone else in the process
treee could "see" it.   This was a "bad thing" (tm) when two people were
compiling ;)

I haven't looked at AIFCHANGELOGON in a while but has this changed?

Also if a process hung you had to kill the whole tree, although didn't
someone say that now the Posix kill command could now handle the
termination of a process in another session's process tree?



Jamie.