On Tue, 29 Apr 2003 Michael.Dillon@radianz.com wrote:
Imagine you have a device that uses lots of addresses but considers them to be sequential numbers rather than bit patterns. For instance, this device could be configured with a starting number and then dole out sequential numbers to connections based on that starting number. This is how a lot of terminal servers work.
Have you configured any terminal/access servers recently?
Imagine that you give the terminal server a number like 223.255.255.200 as the starting number to assign to dialup connections and that terminal server has a 32 port card installed. Then one day an engineer installs a second 32 port card. The terminal server continues to function just fine until one day when it tries to assign 223.255.255.255 to an incoming call followed by assigning 224.0.0.0 to the next call. Suddenly you have all kinds of wierdness breaking out with mysterious broadcast traffic and multicast traffic coming from the device. But it only happens for short bursts during the busiest times of the day. What the heck is going on!?
I'd call that incompetence. A starting number of 200 + 64 ports = too small an IP pool. The cisco gear I use is a bit smarter and when configuring IP pools, both the starting address and ending address are specified (and you can specify multiple non-contiguous ranges). I generally omit /24 network/broadcast addresses from IP pools because too much software assumes everything's a /24 and if you assign someone a /24 broadcast IP, they're going to receive some (maybe alot of) junk traffic depending on what's in the other subnets of the /24 they're in.
Maybe that's why 223.255.255/24 should be forever reserved.
That's way too stupid a reason. That better not be it. ---------------------------------------------------------------------- Jon Lewis *jlewis@lewis.org*| I route System Administrator | therefore you are Atlantic Net | _________ http://www.lewis.org/~jlewis/pgp for PGP public key_________