Re: Identification

[Iljitsch van Beijnum <iljitsch@muada.com>]

On Sun, 16 Mar 2003, David Conrad wrote:

> To me, identifiers are values that uniquely and globally (at least
> within the context of an internet) identify communication end points.

This is old-style thinking.  :-)

What makes a good identifier? Not something like an EUI64 identifier,
for the following reasons:

1. no usable hierarchy, hard to create a distributed database
2. when I replace my network card, I haven't changed my identity
3. It would be extremely hard to make sure each EUI64 is only used once
   globally
4. I could very well be using several network cards but it's still
   always "me"

So we need to identify hosts rather than interfaces. But we can be even
more radical if we want to: an identity could be assigned to a service
or other resource that lives on more than one host. If I want to connect
to www.cnn.com, do I care whether I get www4.cnn.com or www7.cnn.com? On
the other hand, when i have a connection, I want to keep communicating
with the same host for the life time of the connection.

I feel FQDNs are the ideal identifiers. They have a good hierarchy and
the past 10 years or so have shown that building a distributed database
to look up properties associated with them can work well. Also, it is
possible to map arbitrary stuff (IPv4 or IPv6 addresses, phone numbers)
to FQDNs so it should still be possible to adopt "better" identifiers
later. (Or "inferior" ones first for backward compatibility.)

(But maybe I'm getting ahead of things.)

> >  But we cannot hard code identifiers in silicon?

> There is no _technical_ reason this couldn't be done.

Again, how are you going to look them up? Without some form of hierarchy
you need a centralized database. This is unworkable on a global scale.

Iljitsch