Thread Links Date Links
Thread Prev Thread Next Thread Index Date Prev Date Next Date Index

[8023-POEP] Draft 3.1 - Loss of Communication



Colleagues
 
This is an attempt to fix Section 33.7. There are a lot of inconsistencies with the topic "Loss of Communication". There are also multiple comments dealing with loss of communication.
 
Here are the issues that have been pointed out
1) One of the scenarios that lead to loss of communication is when "Management frame communication" is not established after power-on. Loss of communication cannot happen if communication has not been established to begin with.
2) The section speaks about updating MIB variables directly. I think the convention that we follow is that the system sets state variables. The state variables in-turn map into MIB variables. Atleast this is the convention that has been followed in Section 33.6.6.2
3) Multiple senior members feel that power-cycling the PD when LLDP TTL expires might be too drastic a step. We had one comment during the last review cycle and we have at least 3 late comments dealing with this. The recommendation is to get rid of the option to power-cycle the PD based on expiration of TTL timers. The decision to power-cycle must be left to the higher layer protocols/software. We have an umbrella statement that permits the PSE to power-cycle the PD at any time.
4) There might not be a lot of value in sending the Loss of communication information in the TLV between the systems. The PD cannot do much with this information. The PSE could choose to take action if the PD indicates that it has lost communication with the PSE. Not sure how probable is the situation where the link from PD to PSE is working but the link from the PSE to PD is not working. This condition can be detected by the PSE if the PD does not mirror the values back and this condition persists for a long time. We could do away with the "Loss of communication" field in the TLV and make it simpler. We have a couple of late comments that deal with this.
5) The aLostCommunication MIB variables are defined as counters but used as Boolean in section 33.6.6.2
 
I have attached a proposed comprehensive fix for the issues mentioned above. Please go through it and let me know if you have any feedback. If we adopt this then we end up resolving the following comments:
# 145, Late # 10, Late # 5, Late # 16, Late # 12, Late # 2, Late # 9, Late # 8, Late # 11
 
Please treat this as the contribution against my comment # 145
 
Thanks
Anoop

Attachment: vetteth_loss_comms.pdf
Description: vetteth_loss_comms.pdf