[ih] History of Naming on The Internet - is it still relevant?

[Noel Chiappa]

    > From: John Day

    > Saltzer defines "resolve" as in "resolving a name" as "to locate an
    > object in a particular context, given its name."
    > IOW, in computing systems one can't name something without locating and
    > can't locate it without naming it.
    > This is why all of the solutions trying to use it don't scale very
    > well. 
    > What was trying to be done with loc/id split was to find a work around
    > to the fact that Internet lost the internet layer. IP was supposed to
    > be in the internet layer, not the network layer.
    > ...
    > All addresses name the entity in the layer, that they reside. To use
    > Saltzer again, (N)-layer address is a node address, and (N-1)-layer
    > address is a point of attachment address.

I waited to read this until first thing in the morning, when my myalgic
encephalomyelitis has the least effect on my ability to think, but I afraid
that I still do not understand what point you are trying to make.

Let me briefly recount the process I go through in considering this problem,
as my process seems to be very different from the one you are using.

To me, one has to start by ignoring _all_ potential names; one first has to
consider the _kinds of things_ one will need/want to name. I saw/see the need
for two kinds of things (at least, at the levels I was considering; further
up, things like associations might get added, but I'm ignoring them for now).

First, there are the actual things used to carry bits around: physical
links/networks, switches (routers, at the moment), the interfaces to hosts,
etc. Then, we need names for all these things - and the naming system has to
also be able to name (recursively) connected _groups_ of these things - so
that path selection will scale. (Several PhD theses about what to do when
a connected group becomes disconnected ignored.)

Next, there are the fate-sharing entities which are the players in the
end-end communications. I pass over the things one will want to _do_ with the
names for these - which will drive the semantics and syntax of those names.

And, once one has all that, on will need mechanisms to able to relate them
all - e.g. to be able to find the name of the network location (interface) at
which the desired end-end entity is currently to be found.

But, one has to start with the _things_ - the names come later.

	Noel

PS: The "loc/id split" was because a thing (of either of the two classes
above) always has both an identity ('who am I') and a location - but if one
only has a _single_ name to convey both, if the thing changes its location,
one is stuck.