Relay-Version: version B 2.10 5/3/83; site utzoo.UUCP
Path: utzoo!decvax!decwrl!pyramid!hplabs!ucbvax!NPRDC.ARPA!stanonik
From: stanonik@NPRDC.ARPA.UUCP
Newsgroups: mod.protocols.tcp-ip
Subject: 4.3bsd/subnetting
Message-ID: <8606112123.AA02771@nprdc.arpa>
Date: Wed, 11-Jun-86 17:23:00 EDT
Article-I.D.: nprdc.8606112123.AA02771
Posted: Wed Jun 11 17:23:00 1986
Date-Received: Thu, 12-Jun-86 03:30:06 EDT
Sender: daemon@ucbvax.BERKELEY.EDU
Organization: The ARPA Internet
Lines: 25
Approved: tcp-ip@sri-nic.arpa

Now that subnetting (and subnet arp support) is available,
some questions arise about using it, particularly about making
the transition to subnets.

The local network (class B) currently consists of a main cable
to which every host is directly attached.  Because subnetting
was anticipated, addresses were assigned in a subnet fashion;
ie, the third byte (intended to become the subnet number)
grouped machines.  That however seems to cause a problem for
the transition to subnets.  To any subnet branching off the main
cable, the addresses of those not yet capable of subnetting
make the main cable look like a collection of subnets.  Yuck!
4.3bsd doesn't seem to allow more than one subnet per interface
(cable), so many of those hosts will be unreachable.  A nice(?)
solution would be to continue to treat the main cable as unsubnetted
(netmask 0xffff0000) and only route to it packets not bound
for known subnets, but 4.3bsd doesn't seem to do this.

Have we overlooked some routing trick, or must everyone remaining
on the main cable change addresses to one subnet number?

Thanks,

Ron Stanonik
stanonik@nprdc.arpa