The MPLS WG Archive[Date Prev][Date Next][Thread Prev][Thread Next] [Date Index][Thread Index][Author Index][Subject Index] wg last call on draft-heinanen-inarp-uni-01.txt
Juha: Please find below some comments/questions I have on this draft: 1. The document seems to assume the NBMA model, i.e., a set of LSPs is abstracted as a LIS, independent from the IP/MPLS domain which LSPs are established through. I think this should be spelled out. 2. How are the ARP messages encapsulated in case of MPLS LSPs, especially considering their (LSPs') uni-protocol nature? 3. What value is used for the Hardware type field in case of MPLS? 4. How are the *ha fields encoded? Any implications for the case of interface-specific label space? 5. What happens if a request or a reply gets dropped? How often should the request be retransmitted? 6. The scalability section needs more work, I think. It is not enough to say that transmission of the request should be randomly delayed unless you specify the time range. Otherwise we can have a situation where a node is brought up or gets connected to the cloud and every remote node sends exactly one request within a short period of time, but due to the total number of remote nodes we still have O(n^2) messages (requests + replies)? You also need to make sure that even in the situation where the requesters do not behave properly (or are malice) and you receive a lot of requests back to back, reply generation is still controlled... -- Alex |
|