On Mon, 25 Nov 2002, Steven M. Bellovin wrote:
> You run into the birthday paradox here.  In a space of 2^38 address 
> blocks, with 2^19 "allocations" there's a 50% chance of at least one 
> collision.  Given how many home networks will be using this stuff, 
> we'll have far more than 2^19.  
> 
> But that's not the interesting question.  The interesting question is 
> what the odds are of two users of the space "colliding", and that in 
> turn depends on the average connectivity.  On that I have insufficient 
> data.

The requirement for birthday paradox to be valid is when those networks 
are totally interconnected; for that, the number is like 2^19.

I fail to see valid scenarios where even 2^5 site-local networks would be
interconnected.

I'm still assuming the site-locals, are, well, site locals.

So valid interconnections would be:
 1) sites connecting (e.g. two physical locations of one organization)
 2) site-local address info leaking though some ways outside of sites 
where there is site-local connectivity

1) is which seems to be critical, but I still fail to see a huge need for 
interconnection.

2) should not be relevant in the case that collisions are extremely rare, 
as the connections will fail anyway (the question is only _how_).

-- 
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