On Tue, 26 Nov 2002, Michel Py wrote:
> I think that the complete uniqueness is a matter of psychology more than
> anything else. Even if the odds of collisions with a hash or random are
> no better than winning the lottery, there are some people that actually
> win (not me, sigh). Given that Charlie could provide true uniqueness for
> free, why consider a system that does not provide true uniqueness?

I'm not sure whether you haven't been listening or don't agree with my 
(and some others') reasoning.

Nearly unique is IMO much better than completely unique.  (Actually, 2^38 
may even be too unique by the same reasoning.)

That's because those addresses are meant to be unique _only_ within a
scope.  Addresses leak, yes, but that doesn't hurt when the probably of
collision then is suitably low (e.g. below 0.1%).

If the addresses are _unnecessarily_ unique, they will be used for wrong
purposes, no matter what.

It's better that people don't associate _any_ addresses that are not of 
global scope to be globally unique.

IMO we will not want to have site-locals be too attractive: we don't want
to have (too big) competition to global-scope addresses.  IMO we want
_IPv6_ users to use _globals_, and (some) site-locals only when they more
or less have to.  

Globally unique PI addresses are a step in entirely wrong direction.

As you see, these are not really all that "technical" reasons, but I don't
want to see the world where people are using IPv6 with similar
walled-gardens as NAT's create today.

(and besides, with "nearly unique" there's no need for any infrastructure
to support them.)

Note: I'm not disputing the (possible) usefulness of globally unique PI
for e.g.  multihoming etc. purposes, but those are _global scope_
addresses, and that's not the point here.

-- 
Pekka Savola                 "Tell me of difficulties surmounted,
Netcore Oy                   not those you stumble over and fall"
Systems. Networks. Security.  -- Robert Jordan: A Crown of Swords

--------------------------------------------------------------------
IETF IPng Working Group Mailing List
IPng Home Page:                      http://playground.sun.com/ipng
FTP archive:                      ftp://playground.sun.com/pub/ipng
Direct all administrative requests to [EMAIL PROTECTED]
--------------------------------------------------------------------

Reply via email to