Path: utzoo!utgpu!CUNYVM!IBMTCP-L
Date:	Tue, 27 Feb 90 12:43:50 EST
Reply-To: IBM TCP/IP For VM List <IBMTCP-L@CUNYVM.bitnet>
Sender: IBM TCP/IP For VM List <IBMTCP-L@CUNYVM.bitnet>
From:	Andr'e PIRARD <A-PIRARD@BLIULG11.bitnet>
Subject:      Choosing a subnet mask
X-To:         ARPA TCP-IP Discussion Redistribution <TCP-IP@UTDALLAS.BITNET>,
To:	UofToronto LAN redistribution <IBMTCP-L@UTORGPU.uucp>
Message-ID: <90Feb27.134934est.57367@ugw.utcs.utoronto.ca>
Newsgroups: list.ibmtcp-l
Distribution: ut
Approved: devnull@gpu.utcs.toronto.edu

     Choosing a subnet mask is like betting at a race between many
Appletalks vs one Ethernet. It seems that going for too small a number
may have less impact for hosts than networks.
Do these ideas sound correct?
- I could choose a no man's land mask, by this I mean:
- I start with 62 hosts/network, normally a 10/6 split,
- But I use only every 8th network number: 2.0 4.0 etc...
- My split can be described as 7/3/6, I use mask ...255.192,
- If I fall short of networks, I just start using 3.0 5.0 ...
- But I keep away from the odd numbers of larger networks,
- If short of hosts, I may "enlarge" the mask on hosts of that network
only, as long as all gateways on that network do use a per network
mask and the odd number (2.64) is advertised identically to the even
one (2.0) for routing to systems that only know of the smaller mask.
- Should 62 be unrealistic, I may then "enlarge" all masks
progressively.
- It almost impossible to use several subnet numbers on a large local
network, just because large means various systems.

I am assuming static routing information contained exclusively in
gateways for a start.

Thanks for any comment.

Andr'e PIRARD
SEGI Univ. de Li`ege
B26 - Sart Tilman
B-4000 Li`ege 1 (Belgium)
A-PIRARD@BLIULG11 on EARN alias BITNET
A-PIRARD@VM1.EARN-ULG.AC.BE as known to Internet