Path: utzoo!utgpu!news-server.csri.toronto.edu!cs.utexas.edu!usc!apple!oliveb!orc!decwrl!bacchus.pa.dec.com!mogul
From: mogul@wrl.dec.com (Jeffrey Mogul)
Newsgroups: comp.protocols.tcp-ip
Subject: Re: Subnet Broadcasting
Message-ID: <1990Jun13.014727.12840@wrl.dec.com>
Date: 13 Jun 90 01:47:27 GMT
References: <9006121854.AA01650@rp.Wellfleet.Com>
Sender: news@wrl.dec.com (News)
Organization: DEC Western Research
Lines: 22

In article <9006121854.AA01650@rp.Wellfleet.Com> bstreile@wellfleet.com (Bill Streilein) writes:
>Mogul & Postel [RFC 950] specify that the subnet field must
>never be all 0's or all 1's.  Thus,
>
>       128.10.0.0  with a mask of 255.255.255.0 
>
>would be an invalid subnet.  Has the thought on this changed
>since this was proposed?  For instance, if all broadcasts are
>all 1's only, is a 'zero' subnet allowed? 

We recommended against "subnet number 0" because there is some
fear that certain software may use "0" internally to mean something
like "reserved" or "not yet set".  It seemed safer to ban the use
of this value for numbering a real subnet, to avoid any mysterious
bugs.

Also, I'm not sure everyone has stopped using all-zeros broadcasts.

In my experience, it is possible to get away with a subnet numbered
zero, but why tempt fate?

-Jeff