nmanisca@vt.edu (nm) writes:
> At 01:18 PM 8/26/98 -0400, you wrote:
> >
> >>>>>> "nm" == nm  <nmanisca@vt.edu> writes:
> >    nm> o one can never have two seperate networks that each use some
> >    nm> random part of a class c?
> >  
> >  They can use disjoint subnets, but not random "networks" --- if this
> >doesn't make sense to you, then you need to learn a lot more about IP.
> 
> well would you or anyone else like to explain what a disjoint subnet is?

I think what he means is that you can't cut a hole in your network and
route a sub-peice of it somewhere else.  This statement is both right
and wrong. ;-)

The current routing method is called CIDR (classless internet domain
routing).  CIDR basically says that the router should choose the most
specific netmask (smallest network) for all the routing entries that
apply.  Basically you are allowed to cut holes in your network on any
aligned address (eg. an N-address hole has to be aligned to a multiple
of N) and route that hole to a completely different place.  Using a
32-bit IP-v4 address it perfectly acceptable to have 33 "concentric"
networks.

	0.0.0.0/0           -> interface N, gateway #M
	...
	10.1.1.1/32	    -> interface N+32, gateway #M+32

Unfortunately IP over ethernet was formalized long before CIDR came
into vogue.  The ethernet paradyme simply assumes a single flat
fixed-size net.  You just can't punch holes in it and expect everyone
on that ethernet to understand the joke.

If you need 3 ethernets and have a single class-C sized net to play
with you can, for example, break it up into non-overlapping networks
of sizes 128, 64, 64.

-wolfgang
-- 
Wolfgang Rupprecht  		<wolfgang+gnus@spam.free.or.die.wsrcc.com>  
http://www.wsrcc.com/wolfgang/