Subject: Re: RFC-1122
To: Todd Vierling <tv@pobox.com>
From: Erik Bertelsen <erik@sockdev.uni-c.dk>
List: tech-net
Date: 08/15/1997 12:43:55
On Fri, 15 Aug 1997, Todd Vierling wrote:

.. On Fri, 15 Aug 1997, Guenther Grau wrote:
.. 
.. : at work I am "forced" to work with hp-ux ( which is, IMHO, one of the
.. : worst unix implementations I ever worked with. But I don't want to
.. : talk about this right now :-). Starting from version 10.x they forbid
.. : the following netmask: 255.255.255.128 for a class C IP-Address (eg.
.. : 192.13.13.13).
.. 
.. 
.. AFAIK, this "standard" is long since obsolete, and violates CIDR (classless
.. interdomain routing).  In modern interpretation of IP routing, classful (A,
.. B, C) IP interpretation does not exist, and an IP address consists of a
.. network portion and a host portion with an arbitrary netmask from 1 to 32
.. bits in size.
.. 

There exists an RFC (17xx or18xx I think) that clearly states that the
all-one and all-zero subnets are valid, thus implying that netmask
255.255.255.128 is valid giving two subnets, 255.255.255.192 4 usable
subnets, etc. 

I'm sorry that I don't recall the title or exact number.

I used quite a bit of time to investigate this a few months ago, because
we ended up with problems with NetWare servers with this kind of
subnetting.

Since then, Novell has issued patches for NetWare (several versions) that
accept these networks. NetWare 4 accepted the all-zero subnet, but you
still need this patch to use the all-one subnet. 

- Erik Bertelsen