[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: AW: Guidelines for Numbering IPv6 Point-to-Point Links and Easing the Addressing Plans



Tim Chown wrote:
Are there also possibily PIM issues? Would using only link locals cause problems for PIM-SM where the Register messages are sent to the RP from a link-local interface on a CPE acting as the DR? For example, with the Register Stop messages? We ran into this a while ago, but I can't remember the specifics right now :)

The PIM-SM update, draft-ietf-pim-sm-v2-new-11.txt, points out that the source address of the PIM Register message must be reachable throughout the domain. I would assume that implementations would use a global address if the router has at least one global address.

If you use only link-locals on the link, then the CPE hopefully has a global address on the customer side that can be used as source address (or perhaps a loopback address)...

The problem you/we ran into, was that an RFC 1918 address was picked as source address... There might be similar issues with reachability of ULAs perhaps...

Stig


I agree with Fred; from a management point of view, the less explicit numbering required the better, as that implies the least renumbering at a later date...

Tim

On Wed, Mar 01, 2006 at 04:39:30PM -0500, Chip Popoviciu (cpopovic) wrote:

Along these lines there are two other aspects that are worth
considering:

1. A single address with scope larger than link-local would indeed be
sufficient for a router as long as the network management policies and
tools do not need to reach specific interfaces for troubleshooting
purposes for example.

2. In the case of a Service Provider that offers VPN services, an
address with a scope larger than link-local is necessary for each VRF
for things such as PMTU discovery. In that case, it might be easier to
simply assign that address to the PE-CE link rather than create for
example a loopback for each VRF.
In principle link-locals are sufficient for the PE-CE link however in
practice, all little details considered, people might end up assigning
more than that to the link. And in that sense, it will be better to use
a prefix set aside by the SP for its own infrastructure rather than what
is delegated to the customer (as Ole mentioned).

Regards,
Chip

-----Original Message-----
From: owner-v6ops@ops.ietf.org [mailto:owner-v6ops@ops.ietf.org] On
Behalf Of Fred Baker (fred)
Sent: Wednesday, March 01, 2006 11:30 AM
To: Bonness, Olaf
Cc: Ole Troan (otroan); jordi.palet@consulintel.es; v6ops@ops.ietf.org
Subject: Re: AW: Guidelines for Numbering IPv6 Point-to-Point Links and
Easing the Addressing Plans

My assumption as a routing geek is that the only reasonable use of a
numbered point to point link is to address an end station; it needs an
address to be useful in the network. But any router/router point-
to-point interface should be supportable using link-local addresses. A router needs one address reachable by its administration (eg non-
link-local), to be a member of any prefix on a LAN that it is serving,
and to be able to reach p/p-connected end stations it serves.

On Mar 1, 2006, at 5:41 AM, Bonness, Olaf wrote:


  4.  Routing Aggregation of the Point-to-Point Links

  Following this approach and assuming that a shorter prefix is
  typically delegated to a customer, in general a /48 [4], it is
  possible to simplify the routing aggregation of the point-to-point
links. Towards this, the point-to-point link may be numbered using
  the first /64 of a given /48.

using the first (or any) subnet of a larger prefix, breaks the conceptual model of DHCP prefix delegation. the prefix is delegated to the requesting router and cannot be used to number the link between the delegating and requesting router.

My assumption from a service provider point of view would be to use a dedicated sub-preaefix (e.g. /48)of my own aggregate to address the point-to-point links (e.g. /64) to the custumers (in the case I have to do this).

cu
	Olaf