On Mon, 18 Oct 1999, Andrew Gillham wrote:

# Greywolf writes:
# > 
# > "Why should I have to do this!?"
# 
# Perhaps because your are mixing your timeouts?  You're telling a "physical"
# device to drop if it is idle for more than 150 seconds, yet the protocol
# you're running over it, doesn't timeout for (I believe) 14400 seconds.
# Unless you are guaranteed to get the same ip address when you dialup again,
# your open sockets are worthless, and should be closed.

I have a static IP address (demand-connect doesn't handle dynamic
addressing, IIRC).

# Perhaps you should do "sysctl -w net.inet.tcp.keepidle=150", so that any
# socket that is idle for more than 2.5 minutes, is forced to do a keepalive.
# Actually I guess you want a number below 2.5 minutes, if you want the
# keepalive to hold your modem link up.  Then, all sockets need to shutdown
# before your physical link can drop.

Oo!  Didn't know about this one!

# > This "CVS likes to think about things" is bull<bleep>!
# 
# Perhaps so, but do expect an instantaneous response from a freely
# provide service, is a bit much. 

Oh, hey, I didn't mean to bitch about the service -- I think that's fine
and dandy, but if it's "thinking", that's something that ain't working...

# Admittedly I find it disgusting that
# my modem will be completely idle for extended periods.  (over 15 minutes)
# But, this is the nature of the game.  Either use something better, (rsync?)
# or workaround it locally. (keepalives?)

I think I'll try that (sometime) but for right now, sup works just fine.
I might revisit it if/when sup goes away.

Thanks for the suggestions!



				--*greywolf;
--
NetBSD: the power to swerve (penguins, worse than cane toads).