Date: Fri, 14 Jun 2002 09:50:35 -0400
From: Thomas Narten <[EMAIL PROTECTED]>
Message-ID: <[EMAIL PROTECTED]>
| Just because people do something that we feel is unwise, doesn't mean
| we should encourge such behavior (either explicitly, or by
| silence). Saying "don't do that" is one of the things that the IETF
| can do.
Of course. But what here is something that we feel to be unwise.
Use of NAT to make 1918 addresses "work" in v4 is one, but no-one is
suggesting that for v6. Use of 2 faced DNS is another, but while
some are suggesting that (why???) that's also something we should do
without.
But stable addresses ???
| For example, we could arrange to allow sites to obtain globally unique
| address space (though non-routable globally) if they want it for
| internal numbering. That would address some of the issues that private
| addresses of led to. Note that this is something we can't do in IPv4
| because there just aren't enough address for this; IPv6 doesn't have
| that constraint.
Yes, that would be fine, and that's just what Steve Blake suggested in
a message sent a few minute after yours. Stable addressing is what's
needed - the bit pattern that identifies it isn't important.
But note the bit patterns suggested in Steve's message - his suggestion
becomes pretty much the same as the suggestion that I made a long time
ago, to stick identifiers in the upper bits of the SL address (he also
has NUSLA's in there, just to keep everyone happy - and retains current
SL addresses, though I'm not sure why on that) - this suggestion was
thrown out when suggested, the NUSLA suggestion was essentially abandoned,
and every time anyone suggests putting IDs of any kind in those bits, the
suggestion gets laughed at (with no actual technical rationale at all).
So my suspicion is that this kind of approach isn't going to work.
Your suggestion, which is essentially the same as the one I made a long
time ago for all practical purposes, raises the question (or this is what
was put to me) of just who is going to administer this extra address space,
and wouldn't that be a bunch of extra work that no-one wants to do.
At the time (and still) I personally think this is a pathetic excuse for
an argument, but it seemed to be the one that convinced people more than
anything else - fear of creating another ICANN look-like bureaucracy (or
giving that one even more things to make a mess of).
kre
ps: one final comment on Steve's suggestions - I'm not sure that AS
numbers are the ideal thing to use here - for two reasons. One they're
already carried around in the routing system, which might lead some people
to try and use them to make these addresses globally routable, which would
be a big mistake, and second, because they are a routing identifier, I
think it would be best not to confuse things by also making them an
addressing identifier - allocating numbers is such a trivial task that
I think we can just create a new set, or could, if there was any chance of
a proposal like this being adopted, which I doubt.
--------------------------------------------------------------------
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]
--------------------------------------------------------------------
