The point is indeed to specify routing protocol for networks
comprising constrained devices
( e.g. low power) interconnected via lossy links/nodes since indeed
the node failure rate is
likely to be higher than in typical IP networks.
Cheers.
JP.
On Oct 29, 2007, at 8:14 PM, Philip Levis wrote:
On Oct 29, 2007, at 11:30 AM, Kris Pister wrote:
Hmm. You could probably make the claim that there aren't any
wired link layers that don't have losses, too. I guess that it's
all relative. What are the loss rates in wired systems?
The point is relative; losses of course exist in wired systems. A
lot of it boils down to assumptions; if line loss rates are such
that most losses are due to congestion, you want different rate
control than if they are link losses. The many efforts to get TCP
working on wireless, as well as high-speed TCP and fast TCP are
some of the major examples.
Certainly it is possible to find path/channel combinations in
802.15.4 networks that have zero loss at the link layer over
periods of many weeks and millions of packets, which I think gets
called "carrier class" reliability.
It's important to note that this only applies for particular
topologies and environments. Zero loss at the link layer over
periods of weeks is unlikely if you're next to a microwave. It's
the edge cases that always get you.
If we're smart, we'll be able to build networks that use such path/
channel combos almost exclusively.
I think I might word it slightly differently: if we're smart, we'll
be able to build networks that detect such path/channel
combinations when they exist. Assuming they always exist seems
dangerous to me.
Phil
_______________________________________________
6lowpan mailing list
[email protected]
https://www1.ietf.org/mailman/listinfo/6lowpan
