The MPLS WG Archive

[Shen Gangxiang <EGXShen@ntu.edu.sg>]

Hi, folks,

I have some questions on the draft: Some questions on Link Management
Protocol (LMP) Draft-lang-mpls-lmp-01.txt

(Note that if I have missed some information which has been discussed
before, please kindly point out.) 

1. In the Section 4.2.1. Parameter Negotiation
Let's consider the situation that both the local node and the remote node
send the HelloConfig messages at the same time. If so some rules (e.g. the
one according to the value of CCId) are needed to decide which HelloConfig
message will be effective. 

2. In the Section 4.2.3 Control channel switchover
As pointed out, "Control channels may need to be switched as a result of a
control channel failure or for administration purposes (e.g., routine fiber
maintenance, reverting back to a primary control channel, etc.)...." and "To
ensure that both nodes switch to the backup control channel successfully,
both the local and remote nodes MUST transmit messages over both the primary
and backup control channels until the switchover is successful. Messages on
the primary control channel MUST have the ControlChannelSwitchover flag set
to 1 and MUST not increment the TxSeqNum (even upon the receipt of a Hello
message with the current TxSeqNum reflected in the RcvSeqNum field)...". 
I guess that this "control channel switchover" mechanism may not be suitable
for the situation of "control channel failure", because if the primary
control channel fails (e.g. terminated), it may not be able to transmit the
information of  "ControlChannelSwitchover flag". 
4. In Section 4.2.4. Taking a link down administratively
I guess that some more details are required to be looked into if something
wrong happens to the control channel when two nodes are exchanging the
LinkDown=1 message.
3. In Page 10, "When the Test message is detected at a node, .......and
EndVerifyAck message MUST be sent."
As proposed, the component channels are tested one by one. I wonder why we
don't test them at the same time. By doing this, at least we can save the
time consumed by the tests.
Many thanks for comments!
Gangxiang