In your previous mail you wrote:
my bet is that link-local address (for tcp endpoint address) has
the best resistance against renumber or other events, however,
=> this (link-local address idea) has two restrictions:
- it doesn't work with off-link peers (i.e. with iBGP and multi-hop
eBGP)
- it doesn't work if the same address is used on more than a link
(i.e. you have no scope ID to distinguish peers).
there are implementations that cannot do this. also it may conflict
with Francis' BGP4+ RFC (but the RFC is not too clear about separation
between tcp/179 endpoint address, and nexthop values - at least for me)
=> no, BGP4+ is clear and flexible: it addresses only the real issue
(next-hop), not the transport issue (which needs only simple properties
like reachability and local unicity of the peer address).
there also are IXes that use globally-reachable IPv6 prefix, curved
out from one of the adjacent ASes. for example, NSPIXP6 uses
a /64 prefix from WIDE.
=> this discussion occurs one more time yesterday at the RIPE meeting.
There was no general concensus but it seems that global addresses are
strictly better than local scoped addresses. Of course the problem to
get these global addresses was (i.e. is :-) still open. Personally
I think an address has no value by itself so the key is to route
the addresses and a solution is to get addresses from ISPs (IXP members)
which accept to announce/route IXP addresses (i.e. the IXP is a special
case of a multi-homed routing domain). Some others wanted a special block
of independent addresses for IXPs...
Regards
[EMAIL PROTECTED]
--------------------------------------------------------------------
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]
--------------------------------------------------------------------
