The MPLS WG Archive

[Alex Zinin <azinin@cisco.com>]

Yangguang,

> Alex,

> I thought about it. It's actually not true. 

The whole thing is not true or something in particular? ;)

>> I believe it was explained in the thread quite thoroughly already.
>> In short, we have protocols in the generalized MPLS control plane
>> that will not work over a TCP connection, but will require direct
>> connectivity (through a physical link, LSP, or a GRE tunnel) between
>> boxes.
>> 

> Our only concern is OSPF. For example, the transport plane is as below.

[...]

> Optical topology are discovered by LMP and disseminated through OSPF Opaque LSA.
> Even A and D has no direct connectivity in control plane, it still can get
> optical topology through C.

Just in case below ;)

> Wednesday, October 18, 2000, 11:05 AM, Alex Zinin <azinin@cisco.com> wrote:
>> Yangguang,
>> One thing that seems interesting to me in the context of optical
>> networks is potential possibility to prune the topology of IGP
>> adjacencies to minimally necessary (e.g., redundant spanning tree),
>> while still announcing TE information about all optical links.
>> The state of the links would be defined with protocols like LMP
>> or mechanisms integrated into lower layers. This is possible
>> because in case of p2p links (valid for ONs), we do not use
>> IGP topology database for CSPF. However, we do use it in other
>> networks when calculate the paths through multi-access segments.

Note, however, that a) no connectivity between A & B, b) direct
connectivity between A & B and c) a TCP session between A & B
are three "a bit" different things.

Also note, that this quite realistic to imagine a box with gmpls
control plane but without LMP.

Alex.