From: Ralph Droms <[email protected]>
Date: Tue, 18 Aug 2009 09:37:03 -0400
I've had a couple of off-line discussions; now seems like a good time
to ask the WG: does anyone have empirical evidence about the impact of
fragment loss in 802.15.4 networks that would motivate the need for
reliable fragment delivery?
I do not have any empirical evidence on the impact of
fragment loss. I do have empirical evidence on the utility
of end-to-end ACKs for message reliability in a 802.15.4
mesh. The need for end-to-end ACKs stems from the lack of
any reliable self-repair mechanism for mesh routes. When a
route breaks, it isn't enough to wait and resend. Often
some kind of active repair is needed. Speaking figuratively,
when a backhoe cuts a cable, you have to send a truck.
In advocating for Pascal's fragmentation proposal I am
trying to kill four birds with two stones. The birds are:
- using end-to-end ACKs to detect when routes are broken
- avoiding the need to reassemble and refragment large
packets in a route-over network
- increasing the reliability of large, fragmented packets
- compact and efficient source routing
The stones are:
- Pascal's proposal or something like it
- LoWPAN-compatible label switching
-Richard Kelsey
_______________________________________________
6lowpan mailing list
[email protected]
https://www.ietf.org/mailman/listinfo/6lowpan
