Threads and Sockets (Was Ho hum)

[S001GMU at nova.wright.edu]

Date: Mon, 14 Apr 1997 20:12:58 -0700
From: Jeff Kesselman <jeffk at tenetwork.com>

>Um....
>
>You shouldnt be crashing out processes unless you have bad code or a toally
>scrod OS.  Proper definesive programmigjn shoudl handle errors gracefully.
>
>I shudder at the diea thatw e accept code crashing as "normal" or even
>aceptable behavior.

It's not that I consider crashing code as normal or acceptable, I just realize
that these things happen.  No matter how hard you try, you are dealing with 
electronic mediums which are by nature quite volatile and unstable.
There are also times when one would WANT to take down the mud or part of it
for upgrades etc.

Maybe I've just been stuck with diku code for too long and I've never had the
chance to see how a _real_ system would operate... ;)
As far as I can tell, my code is pretty stable, atm.  It doesn't do a lot, but
it doesn't crash either.  *grin*  Maybe after I get into it in more depth and
see that I am capable of writing code that is more stable than DIKU (which
_can't_ be hard) I'll shift the design.  I've put a lot of effort into
modularizing the Message queue communications infrastructure so that it'd 
be easy to yank and replace with something else, if I felt like it

At the very least during developement, when crashing is normal and acceptable
and often desired behavior, it's nice to have the logfile I/O handled
by a seperate process that isn't taken down with the crash and can keep 
logging the info.

-Greg