The MPLS WG Archive

[Neil Jerram <NJ@dataconnection.com>]

The latest LDP MIB draft looks generally good, but I see three problems with the new mplsLdpLspTable, which could be genuine, or could indicate misunderstanding on my part -- so I would appreciate your thoughts on them.  I also have one question.

 

1. It isn't possible to represent an egress or transit LSP (A) at the same time as an ingress LSP (B), if LSP A's incoming (interface, label) are the same as LSP B's outgoing (interface, label), because the index in both cases is the same: (i1, l1).  The old XCMapTable could represent both at the same time, using indices (i1, l1, 0, 0) or (i1, l1, i2, l2) for LSP A and (0, 0, i1, l1) for LSP B.

 

2. There is no mplsLdpLspType value to indicate a liberally retained label.

 

3. In the description mplsLdpLspType, your uses of "ingress" and "egress" (as verbs describing the LSP's relationship to the LSR) are exactly opposite to the conventional usage (which describes a data packet's relationship to an LSP).  This could be rather confusing!

 

4. Do you expect implicit and explicit null labels to appear in this table, and, if so, how?  The point is that implicit and explicit null labels can be distributed multiple times for different FECs in the same session, and it isn't clear in this case how the MIB indexing will work.

 

Best regards,

    Neil

 

Neil Jerram

Network Protocols Group

Data Connection Ltd.

www.dataconnection.com