Date: Mon, 12 Aug 2002 16:05:08 -0400
From: Thomas Narten <[EMAIL PROTECTED]>
Message-ID: <[EMAIL PROTECTED]>
| You seem to be wanting to require that if the u-bit is set, it is
| *guaranteed* to be globally unique. No one can make such a guarantee,
| of course.
No, as I just said in another reply, I'm trying to figure out how the
IID can possibly be used (depending on any uniqueness property) when we
know that that can't happen.
| I have always viewed the u bit a way of indicating that
| the IID is probably unique, with a high degree of likelyhood. This is
| achieved by setting the u bit to 1, iff the ID is derived from an
| IEEE identifier.
Fine. But no-one enforces that. Nothing can rely upon it. I can
trivially create addresses with u set to 1 which aren't created that
way.
In fact, I have just configured one ...
inet6 3ffe:8001:2:181:200::2 prefixlen 64
You can probably ping6 that address. (I hope I got the right magic bit,
I didn't actually calculate, and my mental arithmetic tends to be poor,
the 200 contains the bit I think is 'u').
Now the question here isn't whether or not I'm ignoring the standard,
clearly, as currently written, I am. The question is how is someone
else supposed to tell, how, apart from the fact that the value happens
to be '2' in this case, that this doesn't contain a globally unique IID ?
| I've tried to do so in another note.
Yes, I saw and replied. I'm not convinced, as you will have seen.
| Not so fast. To change the status quo, one needs more than just a
| couple of people saying "yes" and the vast majority of the community
| being silent.
Hmmm...
First, I agree that it is hard to tell what it means when most of a list
is remaining silent. It can be anything between "that is obviously OK,
no-one is arguing against it, nor is anyone likely to, no need to send
a message about that" to "that is obviously bogus, no-one could possibly
agree with that, no need to send a message to point that out".
The usual way, if a proposal has support from more than just one person,
if to assume that it is supported, and say so, and then look for the
reaction. Then if people disagree, there is an obvious incentive for
them to say so, and why - and then there's a better possibility for
a proper disciussion. On the other hand, at that point, silence does
indicate consent - lacking objections, that's about the best mailing
list consensus you'll ever get (of course it rarely happens, someone
always objects, to everything...)
I might also suggest that you contract that position to what was done with
A6. It was the status quo. There there was a debate, with technical
arguments both sides (I disagree with one side, but still, they were
at least real arguments). Quite a few people both ways. And that's
enough for the status quo to change, apparently...
Here, there are still no technical arguments for why any of this is a
good idea - there's the "perhaps some future use" which really should
require at least some kind of suggestion of what it might be, and some
kind of evidence that has some hope of being effective, and the political
nonsense about /48's.
And there's the counter evidence that it isn't implemented, and 2026 which
says "not implemented == remove".
How much simpler can this possibly be?
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 [EMAIL PROTECTED]
--------------------------------------------------------------------
