[ih] "Subnet"

[Noel Chiappa]

    > From: Larry Sheldon <LarrySheldon at cox.net>

    > I do have a question about the term "subnet" as has been used a couple
    > of times.

Do be aware that that term is 'overloaded' - i.e. it has several distinct
meanings. One is the one you list (sort of):

    > as if the subnet number was the third octet in the dotted decimal
    > representation of the IP address, and that decisions could be made
    > using that at the originating end or anywhere in the middle.

Call this 'IP subnetting' to distinguish it. It started as way of splitting an
IP 'network' (originally only class A, since 'IP subnetting' in its earliest
form predates the introduction of class A/B/C IP network numbers) into smaller
pieces, one for each physical LAN at a site. (This is originally an MITism; we
were the first people to do this that I know of.) Actually, at MIT, the subnet
was (and still is, I think) the second byte. Eventually (circa RFC-1122, with
the bit-mask for 'is the destination on my physical network') it came to be
any bit-field in the 'rest' portion. And then CIDR did away with the
network/rest distinction, but that's another generation of development...

The other is:

    > In the ARPANet discussion, what does "subnet" refer to? It seems like
    > there must have been a bigger network that I don't know anything about.

No, nothing to do with that. 'subnet' just means (or, eventually came to
mean) the 'physical network'. In the ARPANet case, the collection of IMPs and
point-point links (since they formed a 'network' of their own, to which the
larger 'network' of all the hosts connected.)


    > I argued and still do that it is mostly a useless datum and can not be
    > known anywhere except on the leg between the destination station ant
    > the router talking to it.

The original concept of 'IP subnetting' was that nobody outside the site
could tell it was subnetted (since originally, subnetting was a purely
MIT-local hack to the architecture).

In the longer term - the discussion of whether routes to pieces of an
naming/topology aggregate can/should usefully be distributed outside of the
boundaries of that aggregate - too large and off-topic discussion for this
margin, alas.

	Noel