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

Re: Draft on Globally Unique IPv6 Local Unicast Addresses

On Tue, 27 May 2003 13:42:19 +0700 Robert Elz <kre@munnari.OZ.AU> wrote:

>     Date:        Mon, 26 May 2003 16:07:25 +0200
>     From:        Brian E Carpenter <brian@hursley.ibm.com>
>     Message-ID:  <3ED21F9D.BC0A3162@hursley.ibm.com>
>   | > > it. But in fact, if I get my copy of "181:8008:7134" from a registry
>   that| > > promises never to give anyone else a copy of it, then it's easy to
>   decide| > > who wins if I ever encounter a clash with another copy of
>   | > > "181:8008:7134".
> I should have asked this when you first said that, but how?
> You (quite correctly) said "a registry".   If you go spend your 10 Euros
> on the number from one registry, and I go spend my 10 Euros from another
> registry, and they both issue us the same number, who wins, and why?

Nobody wins. If the registrys reflect current RIR, then this is not supposed to
happen, the freedom to choose one or another should not permit them to both
allocate the same space, and if they can, the system has a bug. Of course, for a
non-RIR registry, I can't say. But, then, you maybe got a telephone number and
not an Internet address. What did you ask for?

> How/why is one registry to be preferred over another?   By what legal
> fiction would one organisation's right to sell numbers for money be
> able to be judged better than anothers?

In the current legal fiction, there is some top-level co-ordination that
guarantees the field of [0 .. 2**(129-1)] is divided up into non-overlapping
unique spaces (albeit as a sparce matrix, but none-the-less semi-enforced)

To all intents and purposes, ignoring policy differences, 'legally' routable
'integers' cannot overlap.

Of course, there is no (binding) definition of 'legally' here, but by common
convention, the ones coming down from IANA are disjoint spaces. They do not
overlap, and even under flat /32 allocation, they will not overlap unless there
is a BUG in common registry, which will be fixed on being detected.

Now, if you posit some registry outside that agreement (eg, to hypothicate, lets
imagine RF ID tags, or ENUM E.164 mappings), and by some  awful stuffup that
becomes an overlapping number space in the integer field, which has other
behaviour, your question becomes more important.

But, if we stick to "one IANA to bind them" then under current policy, and
predictable grandfather clause "thats not allowed" is a valid answer.

> Note that this is quite different from "routable" numbers, where what
> is really being obtained (sold, leased, loaned, whatever) is the right
> to have the number installed in one particular instance of the global
> routing table - with most people preferring the one that everyone else
> is also using (which does nothing to prevent someone else selling the
> same right wrt some other routing table - that's just not likely to get
> very many customers).

Yes. But one of the things here, is that current number registries go out of
their way NOT to guarantee routability, but do go out of their way to guarantee
uniqueness in the common space. And a non-trivial number of complaints to
registry reflect peoples concerns to keep that true. (eg when ROUTING registries
collide in routing assertions)

> The same is true of domain names, where what is really being "sold" isn't
> the name, but the slot in the DNS zone file.

No, that is different. It *looks* the same in this measure, but you must not
collapse the side effects of a zonefile structured tree, with the integer space.
They aren't the same for all purposes.

> But with numbers that are just numbers and no more than that?

These are not 'just' numbers: we polish the inside arc of the zeroes by hand. 

There are numbers, there are unique numbers, and there are uniquely allocated
numbers. The last kind is what you want I think.

I try to tell my son there really is only one number "three" -If I call it into
use to enumerate three pieces of fruit in the bowl, then some poor sod over in
russia has to forget he has three bottles of vodka, and I don't know how many
lumps of yellow sphere I have once an Icelander needs to count herring in

Not suprisingly, he thinks this is potty. But in Internet addressing space,
there really is only one ::3 I think. And, this is a good thing. 

Or, its not the Internet. 

> 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 majordomo@sunroof.eng.sun.com
> --------------------------------------------------------------------

George Michaelson       |  APNIC
Email: ggm@apnic.net    |  PO Box 2131 Milton QLD 4064
Phone: +61 7 3367 0490  |  Australia
  Fax: +61 7 3367 0482  |  http://www.apnic.net
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