The more important point to me, however, is that the locally generated random prefix is only one of two mechanisms. If I were planning to interconnect a bunch of locally addressed networks, I would get registry-assigned local prefixes anyway -- which are designed to be globally unique. For a nomadic network using locals internally only, the random selection should be fine, even if you interconnect with a small number of other networks on occasion.
--On Friday, August 29, 2003 09:54 +1000 Geoff Huston <[EMAIL PROTECTED]> wrote:
Now if you have a random selection algorithm that truly has 2**40 bits of entropy the chances of any two identified random draws from this space having a clash remains 1 / (2**40), but if you take the total pool of such randomly drawn numbers and ask "how big does the pool need to get to lift the change of a clash within the pool above 0.5?", then the pool size is 1.24 million. If the expectation that this address format is to be used extensively for local connectivity independently of upstream connectivity then I would contend that uniqueness is essential, and the local selection methodology should not be proposed, as its uniqueness properties are flawed.
So I suppose I am saying that this is "nowhere near unique", and that if the entire idea of this concept is to avoid the use of reused private RFC1918-styled addresses and the associated NAT functionality, then this local selection algorithm should not be included in this proposal.
Hans Kruse, Associate Professor J. Warren McClure School of Communication Systems Management Adjunct Associate Professor of Electrical Engineering and Computer Science Ohio University, Athens, OH, 45701 740-593-4891 voice, 740-593-4889 fax -------------------------------------------------------------------- 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] --------------------------------------------------------------------
