[arin-ppml] V6 address allocation policy

[James Hess]

On Sun, Jan 17, 2010 at 9:02 PM, Henry Yen <arin-mail at aegisinfosys.com> wrote:
> On Sun, Jan 17, 2010 at 10:46:03AM -0500, William Herrin wrote:
>> Not exactly... The analysis in RFC 4193 (ULA addressing) section 3.2.3
>> is technically correct but it may be an example of "lies, damn lies
>> and statistics."
>
> Is it shorthand for the Birthday Paradox
> (http://en.wikipedia.org/w/index.php?title=Birthday_problem)?

And the RFC accounts for that,  by using the formula
Exp[-N^2 / 2^41]
http://tools.ietf.org/html/rfc4193#section-3.2.3

However...  I  wonder how  'random'  the 40-bit   global ID will
actually be in practice.

I realize the RFC  suggests a robust procedure for generating it..
But something tells me many sites will  be tempted to ignore those
recommendations,    and  treat ULA  much like they treat RFC-1918
addresses.

That is, they might confuse "random id",  for  just  pick whatever
number occurs to them.    Statistically more than should will probably
 pick all-bits zero,  or  'some convenient numbers'  for the global id
field.

Or  even, a human  picking  the global ID  might  _avoid_  numbers
like  0  or 1234,  such that the number has less than a 1/2^40
chance of being picked.

If the actual  entropy behind  'global id creation'  in practice
turns out to be less than true randomness,  then the results
regarding  probabiliity of collision are also fallible.

--
-J