[4000] in linux-net channel archive
Re: Subnetting, Arghh
daemon@ATHENA.MIT.EDU (Marc Lewis)
Sun Aug 11 12:58:10 1996
Date: Sat, 10 Aug 1996 04:01:58 -0800 (PST)
From: Marc Lewis <marc@blarg.net>
To: Henry W Miller <mill0440@gold.tc.umn.edu>
cc: Mogens Melander <mogens@frontier.dk>,
Linux Net <linux-net@vger.rutgers.edu>
In-Reply-To: <Pine.SOL.3.91.960809065803.4548A-100000@gold.tc.umn.edu>
On Fri, 9 Aug 1996, Henry W Miller wrote:
>
> This depends on your setup. It looks like you want 5 subnets, in which
> case you would need a mask of 255.255.255.31, providing up to 30 hosts on
> each network. This will not fit your scheme, but it is close enough that
> you may wish to make it work. you will have up to 8 different subnets
> avaibal in this case. (32*8=256) if you need more hosts on a network,
> some routers may be able to handle a variable subnetmask, so you could
> make the first 4 networks 255.255.255.31, and the final one
> 255.255.255.127, allowing one big network of 126, and 4 smaller ones of
> 30. (the fisrt network would be x.x.x.0-x.x.x.31, second 32-63,
> 64-95,96-127, and the final one x.x.x.128-x.x.x.255) I'm not sure if
> linux can handle this scheme though, and if it can you may not be able to
> use rip on the network. ospf ought to work, but static routes are the
> easiest way to route.
>
> I woudl recomend the 255.255.255.31 mask, and just stick to 30 hosts per
> network. If you need to you can use this setup:
[snip]
Minor correction to this. The subnet mask should be 255.255.255.224.
255.255.255.31 is not a "valid" subnet mask. The easiest way to figure out
what kind of subnet mask to use is to subtract the number of hosts you want
from 256, so if you wanted 64 hosts on each subnet, your mask would be
255.255.255.192, giving you 4 blocks of 64 addresses each.
FWIW, Linux can handle split subnets within a single class C, so splitting
it up into 2 blocks of 32, a block of 64, then a block of 128 is valid
under Linux, but possibly not some [older] routers.
Some of the older schools of thought also say that you shouldn't use the
first or last subnet in a class C. Apparently, some routers don't like
this very much, but Linux handles it fine.
He is also right about RIP. RIP can't handle subnets, OSPF can, and works
very well under Linux (using gated).
- Marc
-----------------------------+---------------------------------------------
Marc Lewis (marc@blarg.net) | Blarg! Online Services, Inc. - Seattle, WA
Data: 206/441-9109 - or - | Shell - SLIP - PPP - 56k/T1 Frame Relay
Voice: 206/401-9821 | World-Wide-Web: http://www.blarg.net/
-----------------------------+---------------------------------------------
