Shortly after I posted the last one, I realized this problem. I cannot 

just set up a web server (or whatever) to handle the SHUTDOWN signal and 

LOGOFF. If I do, TCP/IP will just bring it back up again! I would at leas
t 
need an interface to say "quit watching me, please!". Otherwise, I need 

both TCP/IP and the server to respond to the SHUTDOWN signal.

Practically, though, web servers aren't a good example. My idea of 
shutting down a web server cleanly is to complete all pending units of 

work (or roll them back) and respond to all user transactions with a 
message "The world is going to end soon!". I don't really want to logoff.
 

I'm on the fence about whether we need a TCP/IP SHUTDOWN. The one example
 
I have seen here of needing it is the behavior of some old communication 

controllers when not properly quiesced. But if its only the old ones, the
n 
then the response to the requirement is "Available".

Does anyone have a real example? Otherwise, it puts this requirement in 

the "Nice to Have" category -- and I won't write it.

On Wed, 9 Aug 2006 17:22:37 -0400, Alan Altmark <[EMAIL PROTECTED]>
 
wrote:

>On Wednesday, 08/09/2006 at 02:28 EST, Alan Ackerman
><[EMAIL PROTECTED]> wrote:
>
>> I'm not sure whether the TCP/IP stack maintains state or not,
>
>It doesn't.
>
>> Just providing SHUTDOWN support for the TCP/IP stack isn't enough,
>either,
>> it would have to pass this along to its guests. (What does TCP/IP do
>when
>> it gets a CP EXT?) Maybe it's just simpler to have each guest register

>for
>> the shutdown signal itself?
>
>There is no way to pass the "shutting down" signal to C/IUCV/REXX/PIPE
>socket applications.
>
>The one thing the TCPIP server perhaps *should* do upon getting a shutdo
wn
>signal is to turn off the listening port monitor.  That would allow the
>apps to do what needs to be done on a case-by-case basis.  No automatic 
CP
>EXT since that could interfere with a "transaction".  I would find it
>highly annoying to have an app close its listening socket, just to have
>the stack force/autolog it.  I'm taking this one under advisement.
>
>Alan Altmark
>z/VM Development
>IBM Endicott
>========================
=========================
========================

Reply via email to