Date: Thu, 01 Mar 2001 11:25:45 -0600
From: "Matt Crawford" <[EMAIL PROTECTED]>
Message-ID: <[EMAIL PROTECTED]>
| awfully close, after all your approximations, to what was given.
I'm not going to start arguing mathematics with my betters, but ...
| There's no use quibbling over a factor of 2 in these very small
| numbers when, as the next paragraph says, this is not at all an
| interesting question, and the real question of interest yields a
| probability far far smaller.
most likely yes - but I realised this morning that the analysis in
that paragraph is flawed.
It isn't sufficient to not have a duplicate with anyone you communicate
with, you also need to not have a duplicate with anyone that any of
them communicate with (or they will see duplicate site-ids at
different sites).
Fortunately, no further steps out are needed, or we'd end up with a
requirement for global uniqueness.
That is, if A communicates with B, and B communicates with C and
C communicates with D, A B and C must all be different, B C and D
must all be different, but it is harmless if A == D.
I'll leave it for those who can do the probability theory to calculate
the risks of collisions for different numbers of peers/site.
But when that is done, do keep in mind that a likely potential use for
this kind of thing is for "groups" of related organisations to set up
a VPN - eg: (say) a fast food chain and all of its franchisees, and its
suppliers - and the suppliers with everyone they supply (etc). The
numbers involved there (even instantaneously, not over an extended period)
can get quite large - 10K is probably not stretching the limits.
kre
--------------------------------------------------------------------
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]
--------------------------------------------------------------------
