Dave Thaler wrote:
> Ffx3::/16 is a subnet-local multicast range. There is no
> subnet-local unicast range today. Charlie is arguing that he would
> like to see fe80::/10 redefined as the "subnet-local" unicast
> range, rather than the link-local unicast range, which means
> that a multilink subnet router (of which a home agent is one
> special case) would be required to proxy ND for link-local addresses.
> (Personally, I think it's way too late to define fe80 as subnet-local.)
>
> Today, a multilink subnet router would have no need to do that,
> other than the problems resulting from anyone not doing DAD on
> every address - which does not appear to be a problem today.
Today there aren't any problems, but that's because nobody
is using very many subnet prefixes on the same {,multi-}link.
I think the current design decision essentially codifies that,
at least for mobile nodes. Is that the intended result?
Another possibility would be to declare that on any link
that has the capability to support mobility agents, a
link-local address semantically has to be same as a subnet-local
address. This would require extra work from the multi-link
proxy agent, but would make the life of network administrators
much easier since my suggested design choice is much simpler
to understand.
In fact, I would argue that the current "architecture" of
link-local addresses as you have detailed it is wrong,
because it's not in the realm of layer-3 design. I'll
have to work out the implications for multicast later, but
my initial impression is that it's workable. I also owe
you more text about why it's a lot of work for home agents.
Regards,
Charlie P.
--------------------------------------------------------------------
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]
--------------------------------------------------------------------
