Padma Pillay-Esnault wrote:

>>Padma Pillay-Esnault wrote:
>>
>>
>>>Acee
>>>
>>>
>>>
>>>>Hi Padma,
>>>>
>>>>I think you can get rid of it. If I remember correctly,
>>>>the grace LSA TLV is solely for the identification
>>>>of neighboring routers on broadcast and NBMA network. In OSPFv3,
>>>>
>>>>
>>>We can derive this information without TLV 3.
>>>Helpers are only for previously full neighbors - which also means
>>>that we know the id/address for each neighbor.
>>>(rfc 2328 - section 10)
>>>
>>
>>Padma,
>>
>>In OSPFv2, OSPF runs on a subnet. So if you have multiple subnets
>>on the same link (e.g., giga-ethernet), you need the local IP address
>>to unambiguously identify the neighbor (or you need to preserve it from
>>the IPv4 packet header). Some implementations don't support running
>>
>
> IMHO, if you do not preserve it from the IPV4 source addr then you have
> a slightly broken implementation ;-) (we do preservation - hence my grunt!)

I agree. In the case of a type 9 LSA, you really need to associated
it with the interface on which it was received in your Link State
database.

>
> BTW, the draft stipulates that one grace lsa is enough to notify the
> neighboring helper even if it has multiple links to the restarting router.

I can't find this in the hitless draft right now - can you give
me pointer?

Thanks,
Acee

>
> Thanks
>
> Padma
>
>
>>OSPF on a secondary address but RFC 2328 allows this and many
>>implementations support it (ours for example).
>>
>>In OSPFv3, OSPF is per-link and neighbors are unambiguously identified
>>by router ID (RFC 2740, section 2.1). Hence, I don't think an IPv6 Interface
>>Address TLV is required.
>>
>>Thanks,
>>
>>Acee
>>
>>
>

--
Acee

• Messages sorted by: [ date ] [ thread ] [ subject ] [ author ]