"ROUTERS" vs. "routers"

[Fred Templin <[e-mail]>]

The v6ops discussion from the 'draft-ietf-v6ops-mech-v2-01.txt' thread
took on an interesting twist that I felt warranted a new subject heading.

RFC 2461 specifies the behavior of traditional routers (i.e., "ROUTERS").
"ROUTERS" typically advertise autoconfig parameters and prefixes from
their attached networks. Hosts use them to reach off-link nodes via default
or more-specific routes. But, a new breed of routers (i.e., "routers") is
emerging from paradigms such as Mobile Ad-hoc Networks. "routers"
typically advertise host routes only (aka, "addresses" or "locators") and
no prefix or autoconfig parameters at all.

I talked about this dichotomy several years ago in my writings
in the MANET wg and in some of the contract work I did at SRI
International. The dichotomy is also discussed in documents such
as the "global6" work by Charles Perkins and his colleagues. But,
the notion of "routers" has really been around for decades and is
quite well known in some circles.

In the MANET paradigm, "routers" often have only a single network
interface which may be used for multi-hop forwarding within a flat
addressing space (i.e., using host routes and redirects to steer the
multi-hop forwarding decisions) while "ROUTERS" are used to reach
other off-link networks. But, the same paradigm can be applied to the
emerging global IPv6 Internet.

In particular, I believe we will soon see an emerging paradigm for
IPv6 in which most (if not all) nodes will be "routers" and a smaller
subset of the nodes will be "ROUTERS". A node can qualify as a
"router" IFF it has the proper credentials so that other nodes can
guard against redirection attacks. The NOID approach specifies
a method by which "routers" may attain the proper credentials
(it also supports site-multihoming as an added attraction):

  http://www.ietf.org/internet-drafts/draft-nordmark-multi6-noid-01.txt

"routers" will also need a means for distributing host routes w/o
including the normal information traditional "ROUTERS" advertise.
The "Default Router Preferences, More-Specific Routers and Load
Sharing" approach specifies the necessary mechanisms:

  
http://www.ietf.org/internet-drafts/draft-ietf-ipv6-router-selection-02.txt

So, what is missing? It would be desireable for a node sending
an RFC 2461 Router Solicitation to be able to specify whether
"ROUTER" or "router" information (or both) was being solicited.
This function can be accommodated by either providing a codepoint
(i.e., 1-2 bits) in the existing Router Solicitation message or a new
IPv6 Neighbor Discovery message type (call it a "Type II Router
Solicitaiton" for lack of a better name.) Perhaps this isn't even
needed; I would be happy to let others comment.

So, where does the "all nodes are routers" model lead us? To an
IPv6 Internet with end-to-end security, site multihoming, multi-hop
forwarding capability, and a restoration of the end-to-end principles.

I think we need to get on board with this, folks...

Fred
[e-mail]



 


--------------------------------------------------------------------
IETF IPv6 working group mailing list
[e-mail]
Administrative Requests: https://www1.ietf.org/mailman/listinfo/ipv6
--------------------------------------------------------------------