Re: draft-ietf-impp-pres-01

Thanos,

My own opinion about the choice:

If an identifier is being used only within a continuing context -- ie,
very short-term usage -- then context-based identifiers are fine.

As soon as the identifier has any complexity of use, either from having
an extended lifetime or from being used across multiple contexts -- even
when they are logically the same context but involve interruptions of
service -- then self-identifying identifiers are much, much better.

The concern about saving bits is usually not worth having dominate the
design choice for such things, in application protocols.

d/

Tuesday, January 14, 2003, 12:31:13 PM, you wrote:
Thanos> I was talking about the difference between context-based uniqueness (what I
Thanos> called the relaxed one - where the context is the watcher/presentity pair)
Thanos> and self-identified (which I think we called "globally unique").

Thanos> In addition to the trade-offs mentioned below, there is also the issue of
Thanos> generating the extra bits for a self-identified unique identifier, as well
Thanos> as the burden of ensuring that this identifier is globally unique.
Thanos> Whoever/whatever is generating those bits must have the means to ensure
Thanos> self-identified uniqueness.

Thanos> Instead of arguing how much that burden is, I was wondering whether there
Thanos> are any reasons to go with a globally unique one, keeping in mind, that
Thanos> there is nothing that prevents us from using a globally unique identifies,
Thanos> in a specific To/From context.  It would still be unique.



d/
-- 
 Dave <mailto:[email protected]>
 Brandenburg InternetWorking <http://www.brandenburg.com>
 t +1.408.246.8253; f +1.408.850.1850




  [reminder: [email protected] for non-technical discussions, please]