Hi Nick,
On Tue, 2003-04-01 at 01:53, Nick 'Sharkey' Moore wrote:
> [Dang! I'd missed this thread under a pile of site-local discussion]
;-)
> So long as all nodes DAD for fe80::X before using Y::X, indeed.
> And it seems to me that it's unsafe to do this any other way,
> since there's plenty of stuff around doing "DIID". On the
> other hand, the 3041 and CGA people are probably wondering why
> they have to check this silly fe80::X first!
Well, they have to do either DAD or DIID anyway. I don't see why they
should care which.
> I myself am happy to go with whichever solution causes least
> harm ...
The safe way, in my opinion, is for all nodes to defend their IDs
against both DAD and DIID, regardless of which approach they use
themselves. I.e., always respond to both fe80::X and P::X.
This way DAD and DIID can happily coexist on a network. Nodes that want
the benefits of a unique ID can do DIID. Nodes that want to share an ID
can do DAD. The DAD vs DIID behaviour could even be configurable per
interface.
> > Perhaps you are suggesting that bad things can happen if
> > nodes do not do DAD, or exhibit some other protocol violation.
>
> Such as Optimistic DAD? I started on draft-moore-ipv6-optimistic-dad
> shortly after Yokohama, where I got the impression that DIID
> was on the way out.
I sure hope not. I don't think we will be changing our implementation,
unless DIID becomes expressly forbidden by a future standards track RFC.
I would be .. somewhat opposed to that.
> It's easily retrofitted to the DIID way of
> thinking, Ed Remmell did this in his implementation.
>
> I'm interested in hearing the thoughts of the group ...
For reasons of efficiency, DIID would seem to make more sense for mobile
nodes.
MikaL
--------------------------------------------------------------------
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]
--------------------------------------------------------------------
