RE: [RAM] Ramblings about "locator"

[<jarno.rajahalme at nsn.com>]

Brian E Carpenter wrote:
>Maybe the essential point is that a locator can at least in 
>principle be mapped to topology and an identifier can't.
>

IMO an ethernet "address" is an identifier. It uniquely identifies the
Ethernet interface in the (name)space of all existing and future
ethernet interfaces.

A locator has some built in topological significance. It is conceivable
that any identifier could be used for a routing and forwarding in a flat
architecture, where there is no aggregation based on topology.

If we had no aggregation in IP addressing, i.e. if we had IP addresses
free of any topological significance (like the ethernet addresses), then
we would not have the current ID/loc split issue either. In a such
system we could do all the multi-homing, mobility, etc. we wanted with
more elaborate routing protocols.

To summarise the point in the ID/loc split discussion is not about
whether something can be used for routing and/or forwarding or not. The
issue is with the built-in topological significance that limits the use
of such "locator" to a somehow restricted topology, when we would want
to have somehow different topology (more dynamic, multi-homed, etc.)
instead. So the issue is about having identifiers free of network
topological significance. It is yet another discussion whether these
identifiers could/should have semantics on some other axis, e.g. whether
these identifiers should be somehow structured for organizational
(=administration/"ownership") reasons.

Regards,

	Jarno Rajahalme

_______________________________________________
RAM mailing list
RAM at iab.org
https://www1.ietf.org/mailman/listinfo/ram