Re: [RAM] some draft proposed definitions

[Mikko Särelä <mikko.sarela at ericsson.com>]

Here's my take on some of the definitions.

Identifier: An element of a name space that persistently and uniquely
identifies an entity, e.g. independent of the location of the entity.

Locator: A string that indicates a particular point in the network
topology.

Identification: A means that an entity uses for identifying another
entity, e.g. TCP uses a Locator (and port number) to identify the other
end point.

Address: A locator that is used for identification purposes.

Note that identifiers can be used for forwarding data (e.g. ad hoc
network routing protocols typically work with identifiers that have no
topological meaning) and locators can be used for identification
purposes (as TCP and UDP do).

To me the question seems to be whether the naming should be based on the
usage of an object, or by the purpose for which it was created. In the
first case, you can say that if somebody uses a locator for
identification purposes, then it is also an identifier. In the latter
case, you would name the object based on its primary purpose (i.e. an IP
address would be a locator used for identification purposes and not an
identifier).

If one does go for the former approach, then we will need to have
separate definitions for persistent and non-persistent identifiers and
possibly others, whenever somebody decides to use a new class of objects
for identification. Similarly, if we define a locator based on the idea
that it is used for forwarding, we will have to accept that we will have
topologically non-meaningful locators, i.e. in ad hoc networks it is
common to use flat identifier space for routing. And then we will need
to define topologically meaningful locator and topologically
non-meaningful locator separately.

I think this approach is doomed to unnecessary complexity. Instead, I
would go for the latter approach, and name objects based on their
qualities and then denote the usage separately (i.e. IP address is a
locator used for identification rather than an identifier).

-Mikko


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