On Wed 07 Feb 2001 at 09:50:32 +0900, itojun@iijlab.net wrote:
> >>I don't think this is desirable (and IPv4 does not do it like this). Is
> >>there any spec about it?
> >	source address selection algorithm for IPv6 is documented in
> >	sys/netinet6/IMPLEMENTATION, or
> >	http://www.kame.net/dev/cvsweb.cgi/kame/IMPLEMENTATION.
> >	(latter one is more recent, but may have some changes made after
> >	the current integration).

3ffe:8114:1000:0000:0000:0000:0000:00dd ipng.nl tunnel		I
3ffe:0b00:0c18:1fff:0000:0000:0000:0149	freenet6 tunnel		
3ffe:2a00:0100:3001:0000:0000:0000:0002 anoncvs.no.netbsd.org	D

It is true that the prexix match between anoncvs.no.netbsd.org and the
freenet6 tunnel is slightly longer than with the ipng.nl tunnel.
This provides the disambiguation for step 1:

  1. If nonDEP(samescope(D)) is not empty,
     choose a longest matching address against D. If more than one
     address is longest matching, choose arbitrary one provided that
     an address on I is always preferred.

However, this difference is so small that I would say it should not be
considered significant.

> 	there's a spec about it (draft-ietf-ipngwg-default-addr-select-02.txt)
> 	but NetBSD/KAME code is not doing this (yet).

I will read that document next.

> itojun
-Olaf.
-- 
___ Olaf 'Rhialto' Seibert - rhialto@polder --Soep van de dag, wat zal dat zijn
\X/ land.nl     --wat kan dat wezen, beter maar het ergste vrezen -Boy Bensdorp