[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: Misusing registries for uniqueness (was Re: Draft on Globally Unique IPv6 Local Unicast Addresses)

    Date:        Fri, 30 May 2003 16:01:15 +1000
    From:        Greg Daley <greg.daley@eng.monash.edu.au>
    Message-ID:  <3ED6F3AB.6050003@eng.monash.edu.au>

  | It's not that dumb an idea, it reminds me of
  | base-85 (RFC-1924) IPv6 addressing notation.

Which is a joke, not an idea (dumb or otherwise)...

  | It certainly does solve the uniqueness issue
  | for any given instant.  What happens if you
  | 're-number' your telephone?

This has the same problem as MAC addresses - it isn't stable (someone
else can quite validly end up with the number that used to be yours).

But, if abusing number spaces to try and gain easier uniqueness is
an aim, then a (less automatic but perhaps still reasonable) method
would be to abuse the IEEE OUI space (instead of the MAC address space).

That is, any organisation with an OUI (22 real bits big) has 18 bits
(256K) of numbers they can allocate however they see fit.   And of
course, organisations that the IEEE allows can go get new numbers.
(The IETF is one such organisation of course, so IANA would have
numbers to allocate).

(This re-use of numbers would not conflict with other uses of the OUI,
it would be in parallel, just as MAC48 and EUI-48 are - it would be
xxx-40 of course, or -45 or something, depending upon prefix length).

Of course, IEEE would probably need to agree to their number space
being abused this way before we could suggest it as a method.

Doing this still doesn't guarantee any kind of uniqueness, all it does
is provide a ready made answer to the "one monopoly organisation"
problem, while similtaneously making it effectively impossible for
anyone to snarf any large fraction of the address space (no need to
prescribe huge fees - like Eur 10).

Even adding in registries, as Benny Amorsen proposed, doesn't really
solve the problems.  To be unique, there'd need to be one overall
registry (perhaps sub-dividing the space to others) - which again means
one monopoly, with no real justification for anyone to be the owner of
that.   Further, it only helps if people actually register (otherwise
there are the same issues as with trade marks, where they can be held,
unregistered, for a very long time, etc - except we have no legislation
or case law to help with consistent reasonable answers to who wins).

Multiple registries just means more places to register, and would create
more confusion if A registers in registry 1, and B registers the same
number in registry 2 (perhaps at about the same time, perhaps not).

To actually get any kind of uniqueness, we need something that fails to
work if the numbers are duplicated anywhere - and as these numbers are
addresses, that really means that addressing has to fail.   I can't see
how that can possibly happen for numbers intended to be used only locally.


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 majordomo@sunroof.eng.sun.com