Getting back to the original question:

> We have about fifteen printers configured to run on the network. We
> are running out of IP addresses so the proposal has been made to add
> new sub-masks. The printers would be put on there own sub-mask. They
> currently reside on the only (help me here, gateway?, mask?, path?)
> along with the CPU. If we move the printers to a new sub-mask what
> do I need to do?

If these printers do not need to be accessed from the Internet, just
move them to a private subnet in RFC1918 space (10.x.x.x, 172.16-31.x.x.
or 192.168.x.x) and you can use them internally.  Assuming you have a
router (in NMMGR it would be the default gateway)
you have the choice of placing the printers on a separate subnet (a new
interface of the router) or giving the current interface a secondary
address.  For printers, 192.168.1.1 would be fine for the router
interface address, and your subnet mask is 255.255.255.0.
Just DO NOT advertise these addresses into the public internet.

As for the subnet discussion, for the most part the old class A,B,C
addresses have broken down to more generic CIDR (Classless Internet
Domain Routing) blocks.  Every IP address has a "network" and a "host"
component.  With classful addressing, there were:

    126 class A networks of 16777216 hosts each,
  16382 class B networks of    65534 hosts each, and
2097150 class C networks of      254 hosts each.

This depends on the "bits" allocated to designate class, network, and
host; plus you aren't supposed to use the "zero" host address, nor the
maximum (all bits ones) host address as it is the broadcast address.
"Subnetting" started by subdividing the existing classful networks into
smaller segments, such that a classic class C address could be divided
into:

      2 subnets of 62 hosts,
      6 subnets of 30 hosts,
     14 subnets of 14 hosts,
     30 subnets of  6 hosts,
     62 subnets of  2 hosts.

The "excess" of the possible 256 are used up by the old standard:

   First and last subnet are reserved,
   First and last host number are reserved.

Initially, subnetting had to be consistent, and not all routing
protocols could handle subnet advertisements at all (RIPv1 for
example).  Then came VLSM (variable length subnet masks) where you
could mix sizes/lengths/prefixes of subnets, but it required advanced
routing protocols (RIPv2, OSPF, EIGRP) to support.

CIDR relaxes these rules in that you can use "subnet zero" but the last
subnet is still somewhat sacred, being the broadcast address.

CIDR addresses are given as "prefix/length" as someone alluded to
earlier.  A classic class C might be 192.168.1.0/24 meaning that the
network portion of the address is 24 bits long, leaving 8 for the host.
The "subnet calculator" pointers allow you to easily calculate
prefixes, masks, and broadcast addresses.

In the "bigger picture" most ISPs are granted large, contiguous blocks
of addresses, on nice binary boundaries; these in turn are leased out to
their customers.  This enables the ISP to advertise routes to the global
internet with the aggregate prefix.  The global internet community (at
least the admins) are striving to keep prefixes as short
as possible to minimize the size of the global routing table.

But enough for now, class dismissed :-)

Jeff Kell <[log in to unmask]>