RE: FCIP: draft-weber-fcip-encaps-00.txt

["Sriram Rupanagunta" <sriramr@aarohi-inc.com>]

Few comments on the new proposal for FCIP changes:

- Section 3.2: It is not entirely clear how the FCIP devices would
synchronize
clocks across the IP network. There is mention of using NTP or SNTP - would
that mean they operate independent of Fibre channel time services, operating
as a IP service ? I would imagine this to be the case, but not sure if this
is mentioned clearly enough.

- Section 3.2: The mention of time stamps for time out error recovery
needs to be more specific, IMO. It says that "if too much time elapses,
the FC frame should be dropped", which is not precise. I believe it should
specify how it relates to the two FC timeouts.

- Section 5.1: The reason for additional checksum is not clear enough.
If this is solely for the purpose of re-synchronization, would it
justify the additional overhead for every FCIP frame ?

- Both the original ID and this proposal leave out the part about
aborted TCP connections, but talk about re-synchronization during data
transfer phase, in the face of packet loss, congestion etc.
I would imagine the re-sync to be more critical "after" a connection
is reset, since there is data in transit and in the receiver queue.
Do the authors believe that the new proposal would be sufficient
to address the TCP connection resets ?

Regards,
Sriram