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

Re: Draft on Globally Unique IPv6 Local Unicast Addresses



    Date:        Wed, 28 May 2003 22:42:52 +1000
    From:        George Michaelson <ggm@apnic.net>
    Message-ID:  <20030528224252.471dc24f.ggm@apnic.net>

  | Strong admission checks into routing are going to make Joe's numbers
  | less useful.

Huh?   My numbers are never going anywhere near anyone's admission
checks, except mine.   If I want to communicate with the world, I will
have, and need, regular global (routable) addresses.

The issue here is not that my numbers aren't unique, as I expressly don't
care, but that someone else thought they were getting a unique number, and
paid for that, and it turns out that their number is just the same as the
one I'm using - that is, they didn't get what they paid for at all.

If you're promising to sell duck eggs, duck eggs better be what you're
providing, just hoping that the eggs you supply never hatch, and so no-one
ever finds out they were emu eggs isn't good enough.

  | If you want to deliberately go and break the global routing cloud you
  | can do that. Sooner or later, you wil be caught. People even went and showed
  | they could run alternate roots in DNS. But your punters, people who bought a
  | thousand from Joe for $10 instead of leasing one from an RIR for $100,
  | only do that once, if they fall into a hole fast enough.

Yes, that's fine - provided that there is something that doesn't work
for the "bogus" number.   Here, for almost everyone, there's nothing at
all that doesn't work for any number they simply invent.

Maybe major corporations will want numbers that don't conflict with
those of other major corporations, just in case they ever do want to
use the things for direct communication.   They'll probably pay whatever
fee is involved.   That's fine.   Just don't try and tell them that the
numbers allocated are "unique", because they're not - except with the
domain of end users who obtained their numbers from the same base source.

For us, as protocol designers, what this means is that no matter what
we do, nothing, anywhere in this space, can ever be assumed to be a
unique number.   In some operational environments the number is likely to
be unique (no-one else who matters is using the same one), which is fine.
But the protocols can *never* assume that.

  | In this matter, there are no bargains. If you seek to avoid or buck
  | the system, you will pay a cost, in some measure.

How?   Exactly what cost am I (on my private network at home) ever going
to suffer because of making up my own number instead of paying for one ?

  | Yes. And, if you select a suitable prefix, I don't see any reason not
  | to say "this prefix is for sites to use random selection methods to
  | get 'nearly unique' addresses, if you want that. People who don't want
  | to see this can filter it!

I expect all of this to be filtered just about everywhere.   Explicit VPN
type setups perhaps excepted (though even there, there's no requirement to
use these non-routable addresses - the sites will also have their global
addresses that they can use, a VPN after all is really just more trusted,
or less filtered, traffic that is known to come from a particular source,
it doesn't have to be using private addressing, and can be simpler if it
isn't).

  | No. this is like 'relatively prime' -if you are *reductionist* then you
  | can't run an Internet anyway. one exception doesn't break 'practical'
  | uniqueness.

If it is the right exception it most certainly does.

  | We've been living with 'good enough' for a long time, in a lot of different
  | fields, not just Internet/Networking. How many more than one does it take to
  | make uniqueness practically useless? I think a HELL of a lot more.

George, the problem is that if we claim that the things are unique, people
are going to start relying upon that.   That is, building protocols that only
work if the numbers are unique.  Or, in other words, one of the problems with
site-local addresses as we have them now is that they're not unique.   This
proposal is supposedly fixing that.   Except it isn't, and we must not fool
ourselves into believing that it is.   Nothing that had problems because
SL addresses weren't guaranteed unique will lave less problems with this
proposal.

That's fine as far as I'm concerned.

  | its not about single point decisions. If 2 reasons exist making it useful
  | isn't that enough? When you argue against one, does the other disappear?

Once again, it isn't that unique numbers aren't useful, it is that we have
no way to make such things.

  | >   | Plus, the ability to defer a decision to change ones mind later on.
  | > 
  | > Change one's mind about what?
  | 
  | About needing uniqueness. You can leave the egg whole and break it later,
  | but its harder to glue it back together if you break it first and change
  | your mind...

That's fine, provided uniqueness exists to begin with.   Here we have it only
if everyone agrees to play by the rules, and do the right thing.   Everyone.

Do you really believe that can possibly happen?   Remember the proposal here
is to have some entity selling up (up to) 2^40 numbers at 10 Eur each.
That's 10,995,116,277,760 Euros (11 US style trillion) available.
And the cost of obtaining the raw materials - 0 Euros.   The only costs are
the distribution costs (the cost of receiving and sending a packet per
number, and a system upon which to run it all), and the billing costs.

We've already seen people in the name space attempting to (and to a small
degree, succeeding) sell names that patently obviously couldn't possibly
work - do you really believe that there won't be similar people selling
numbers, that patently obviously work just fine (all that's needed is that
me and all my friends get our numbers from the same group if we plan on
using them between us)?


Tony Hain <alh-ietf@tndh.net> said:
  | I believe the root of kre's concern is that we don't approach the governance
  | space with the appropriate attitude. We need to admit up front that numbers
  | will never be absolutely unique, and that some people will want to make up
  | their own for completely random reasons.

Yes.  "We need to admit up front that numbers will never be absolutely unique"
Exactly.

But ...

  | All we can do is define a single rooted registry with a to-be-defined
  | conflict resolution process

I'm not sure this is possible (I don't mean technically possible, I mean
politically possible).   From where to we get the power to decide who is
the ultimate authority for numbers ?    As George pointed out, even for
the current (IPv4) routable numbers, the registries don't attempt to
promise that they're useful - it just happens by common consensus that
they are, but if one of the registries allocates you a number, and for
whatever reason it doesn't work, there's nothing really that you can do
about it (if you're really lucky you may get them to refund the fee).

Here, there's nothing requiring uniqueness for the numbers to be useful
(except for those few organisations that want VPNs between them, for which
the only requirement anyway is that none of the organisations involved are
using, or seeing, the same numbers).    I can imagine a VPN with a few
thousand members, much bigger than that, and I suspect that what you have
is just a public internet (that is, with all the same problems, of wanting
to be using topological addressing to make the routing scale sensibly, and
hence not wanting to distribute around "non-routable" numbers anyway).

Brian said that he wanted to avoid another ICANN nightmare (my words, not his).
He didn't say how he planned on doing that.   With 11 trillion Euros
available for someone (over time) there's sure going to be a lot of pressure,
both from organisations that want to get their finger in, and even from
governments who want all of that foreign exchange flowing into their country,
rather that flowing out to some other.   Is the intent here to just slip
this through and hope that no-one notices?

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
--------------------------------------------------------------------