[KLUG Members] Why this routing, anyone?

Adam Williams members@kalamazoolinux.org
Mon, 01 Dec 2003 17:50:17 -0500


> I think this would alleviate some of my problems at work. What this says to me
> is that RH realised that <blank> happens and sometimes not all computers get

RH didn't "realise" anything.  This is what is now supposed to happen, 
zeroconf is a standard used by a great variety of consumer devices
including things like the Apple AirPort and a variety of handheld
wireless devices.  

> assigned a DHCP address*. This will route all traffic from those computers to
> the real network, if I am reading the line right [I've never put enough work
> into routeing]. Though, I would rather delete it and know when a problem occurs.

No, it won't route to the "real" network.  ALL routers are supposed to
drop any and all 169.254. packets.  This subnet is not routable. 
NAT/Masq applications are supposed to drop any and all 169.254.
packets.  If you need Internet access you need to setup real
networking,  otherwise zeroconf could create inadvertant holes in a
network.  

> *This seems to happen very often to our 95 and 98 dinosaurs. When it does,
> MS gives them a 169.254... address, unfortunatly it usually conflicts with
> one of the other computers who have failed to get or refresh the IP
> address.

Sure, we see the same thing.  Zeroconf only makes a really crappy
IP/DHCP implementation even more annoying.  It can be disabled in Linux
(RedHat at least anyway) and Mac OS/X.  I've never seen a document about
how to disable it in any version of Windows (although I'd wager there is
some obscure registry key).