Re: about Area Splitting

[Acee Lindem <[e-mail]>]

Li, Ke Qin (Peter) wrote:

>Hi,
>
>Thank you for your reply.
>
>Let's still use the notations in that draft. In my understanding, when a
>packet destined to y reaches a, a will discard the packet, and reply an ICMP
>Destination Unreachable message. But this will not change the routing table
>of c. C will keep sending packets to a.
>
Correct.

>
>What does "reject route" mean? And what does "coincide with computed
>intra-area routes" mean? Do you mean that because m is not equal to x, a
>reject route corresponding to m-x will be installed?
>
When you advertise a prefix you should install a corresponding route (if
there is no intra-area prefix that corresponds
exactly to the area range). Hence, you install a route with a null
nexthop and reply with an ICMP destination unreachable (with
normal ICMP rate limiting). I believe the "reject" terminology dates
back to BSD 4.x although others may correct me if
they are aware of a different origin.


>
>Thank you.
>
>Keqin Li
>
>
>
>>-----Original Message-----
>>From: Acee Lindem [mailto:[e-mail]]
>>Sent: Tuesday, December 09, 2003 10:38
>>To: [e-mail]
>>Subject: Re: about Area Splitting
>>
>>
>>
>>Hi Kegin,
>>
>>The traffic should result in an ICMP reject since any deployable
>>implementation will install a reject route corresponding
>>to its active area ranges that do not coincide with computed
>>intra-area
>>routes. RFC 2328 doesn't state this explicitly but
>>it is implied by section 3.5 (it says a single route may be advertised
>>for an area range - this implies a route exists).
>>
>>I'm not sure how serious the problem is since there haven't been that
>>many complaints. However, as
>>Sina points out, the proposed Tunnel Adjacency (TA) extension can be
>>used to repair a partitioned
>>non-backbone area.
>>
>>
>>
>>
>>
>
>
>