Thanks to:
"Otter, Richard" <rotter@haley.genetics.bio-rad.com>
Juergen Wagner <gandalf@Csli.Stanford.EDU>
anderson@neon.mitre.org (Mark S. Anderson)
bern@penthesilea.uni-trier.de (Jochen Bern)
mwebb@gaea.erin.gov.au (Mike Webb)
rstar@unxsscivan066.sierrasys.com
syd@dsinc.myxa.com (Syd Weinstein)
Thanks to Syd and Juergen for very complete explanations and some
excellent suggestions on how to do this properly. Thanks to Mark
for the most concise reply :-)
In short, when you subnet networks, having the subnet *network* portion
as all 0 or all 1 is a no-no, it really is used for broadcasting to all
the subnets. I.e. ff.ff.ff.80 is impossible, and in ff.ff.ff.f0, you
can't use X.X.X.00 or X.X.X.f0 as network numbers.
That's the way it is, like it or not. It does work on some systems, as
they don't implement the broadcast mechanisms that this is suppose to
represent. Apparently the motivation for subnetting (hearsay follows)
was to allow the efficient use of class B networks when physically
subnetting. C# or D were meant for very small networks, and breaking
a C in 2 or 4 wasn't really a consideration. The broadcast mechanism
is therefore used so that breaking up a network physically up doesn't
cost you in terms of things like boot servers needing to be replicated.
Suggestions on how to do what I want were:
1) use DHCP - it will dynamically allocate IP numbers to hosts as they
are required. If you have a small number of simultaneous IP hosts,
then you only use a small number of IP addresses. Gets harder as
the number of IP hosts approaches the number of addresses, and you
may have machines reallocating all the time in our situation.
2) use an unconnected class B address and firewall/proxy services. There
are apparently a number of class B addresses that are allocated to be used
for this purpose. Internally, J random host connects to proxy services
on your firewall/proxy machine, and *that* machine and address are what
the internet sees, so it doesn't really matter what your internal address
is. The same network numbers can be re-used and save address space, but
you better not let one leak out (not a problem, as most Internet providers
will screen the packet anyway).
We are probably going to go with the latter method. Interestingly enough,
most of the local folks I talked to came up with the two solutions for
other reasons, but not many people know about the broadcast subnet bit.
l & h,
kev
Kevin Sheehan kevin@uniq.com.au |
Uniq Professional Services Pty Ltd ACN 056 279 335 | Why Not?
PO Box 70, Paddington, NSW 2021, (Sydney) Australia |
W: +61-2-380-6360 Fax: +61-2-380-6416 Pager: +61-16-287-000 |
This archive was generated by hypermail 2.1.2 : Fri Sep 28 2001 - 23:10:16 CDT