On 3/16/2016 4:57 AM, [email protected] wrote:
Hi, Eric and all,
Try to answer your questions below, please comment.
Q: If the core runs only IPv6/MPLS, and the next hop of a 4PE route is
an IPv4 address, I don't really see how the next hop resolution is
going to work, as the next hop (a v4 address) will not appear to be
reachable through the v6 core.
A: No, you can not use the v4 next hop to reach the egress 4PE in the
v6 core directly. But, the ingress 4PE can use the v4 next hop to get
the corresponding v6 address of the egress 4PE and forward the packet
toward the egress 4PE through the v6 LSP after encapsulation.
Encapsulation here means adding two labels. Ingress 4PE needs a data
structure to establish the connection between the v4 next hop and the
the corresponding v6 address of the egress 4PE. This data structure is
implementation specific.
I think the key statement here is "Ingress 4PE needs a data structure to
establish the connection between the v4 next hop and the the
corresponding v6 address of the egress 4PE. This data structure is
implementation specific." Whatever implementation technique you choose
to use, you are still treating the v6 next hop as the real next hop. I
don't see why the protocol should not carry the v6 next hop in the next
hop field, per RFC 5549.
Q: I think your proposal really is intended to treat the v6 address in
the NLRI as the next hop. But that leaves open the question of why
you want to put the next hop address in the NLRI field instead of in
the next hop field.
A: What I want is to send both the IPv4 and IPv6 addresses of the 4PE
to other 4PEs.
But the IPv4 address of the 4PE router plays no role in the protocol.
In the implementation, you seem to be using it only as a sort of pointer
to the IPv6 address of the 4PE router.
Q: Some people think it's a bad idea for the prefix and the next hop
to be of different address families; those folks tend to regard RFC
5549 as a bad solution.
A: I don't think I am one of those folks.
Q: However, I don't see what advantage your proposal has over RFC
5549. In order to determine whether a given 4PE route is feasible,
or whether it is the bestpath, you still have to resolve the IPv6 next
hop, you still have to consider the IGP distance to the IPv6 next hop,
etc.
A: If 4PE only gets the v6 addresses of the other 4PEs, how does the
4PE build its IPv4 routing table? Can it install a v6 address in its
v4 routing table?
Whether an IPv6 address can be installed as a next hop in the IPv4
routing table is an implementation issue.
I don't see any reason for the protocol to carry around the IPv4 address
of the 4PE router, as this information doesn't really play any role.
If you put the next hop in the NLRI, and put a meaningless value in the
next hop field, you have to revise all the bestpath selection rules.
_______________________________________________
BESS mailing list
[email protected]
https://www.ietf.org/mailman/listinfo/bess
