[DNA] Re: The difference between identity and identifier (was Re: DNA GoalsIssue List)

[Erik Nordmark <Erik.Nordmark@sun.com>]

> Well, independent of the real technical argument (where I lean
> towards Erik's argumentation), I think that there is a lurking
> language issue here.  According to my experience, people seem
> to often have a difficulty in making the subtle but important
> difference between "identity" and "identifier".

Sorry for being sloppy.

But I think in the case of "link" we have a harder problem as well,
which is that I sure can't point at the identity of a link.
For a host I can decide that the identity is tied to some physical
aspect of a device, or some set of bits in a memory chip.
Similarely for a process.

But philisophically, what is a link?
It isn't an Ethernet switch, or any single physical object.
And it isn't a set of bits like a storage/disk volume, or a process.

Instead the identity itself feels very amorphous.
When a link is partitioned by switch or wire failures, does it stay as
one link or does it become two?
If it stays as one, how is this different than two links that are physically
adjacent? Should the network admin decide that the partitioned link
was a good thing and not repair the switch/wire, but instead assign separate
subnet prefixes to the two halves, then at what point in time does the single
link become two links.

For these reasons I'm having a hard time pinpointing the identity of the 
link, without talking about the identifiers that have been assigned to it
by something external to the link object.

   Erik