For some systems, being able to craft unique identifiers is helpful.
For some of those systems, mandating and attempting to guarantee that
uniqueness is also helpful.
For other systems, it is not necessary.
And when it is needed, there are multiple possible ways to craft such
things. Any technique wew use has some risk of failure. The risks are
in different places. (Vendor manufacturing, allocator error, ...)
It is not the case that we need to define one such mechanism, and
mandate that it be used by all solutions. The overall problem space is
constrained in many ways, and is mult-dimensional. It is fair to
criticise a solution if it requires a specific property and then
provides mechanisms which do not sufficiently ensure that property. But
the proposed solution gets to pick the mechanism, define the degree of
confidence required for the system to work, and then put forward an
answer.
Domain names are sometimes a useful answer. For example, they are
propsoed for the name based socket work. They have been discussed in
other contexts at times. They have benefits, and costs.
Yours,
Joel M. Halpern
Toni Stoev wrote:
On Saturday 17 July 2010 at 20:36:16 [email protected] sent:
Toni,
I think you are pursuing the wrong objective (which is global uniqueness of the
identifier).
Does anyone else think global/universal uniqueness of the identifier is the
wrong objective?
...
Regards
Toni
_______________________________________________
rrg mailing list
[email protected]
http://www.irtf.org/mailman/listinfo/rrg
