On May 27, 2008, at 2:29 AM, Carles Gomez Montenegro wrote:
>
> At least, there are the following items listed in the routing  
> requirements
> draft (http://tools.ietf.org/id/draft-dokaspar-6lowpan-routreq-05.txt)
> that route-over approach cannot provide or would provide only in a  
> limited
> way:

I disagree wholeheartedly. I think there is a big confusion here  
between a protocol specification and a protocol implementation.

These are all arguments for cross-layer design, that tightly  
integrating routing and the link layer will lead to a better solution.  
Practice has shown us otherwise; only by clearly separating concerns  
do you give a protocol the flexibility needed to evolve over time.  
What happens when a new low power link layer emerges? Having N  
different solutions, each with their own details, which somehow need  
to be made to work well together, seems like a path of brittle and  
difficult to manage networks. Switches are good, to a point; there's a  
reason you have routers.

The purpose of a protocol specification, IMO, is to specify what is  
needed for interoperability; there should, in theory, be multiple  
possible implementations (and by that, I mean power saving approaches)  
that can all interoperate.

Phil

_______________________________________________
6lowpan mailing list
[email protected]
https://www.ietf.org/mailman/listinfo/6lowpan

Reply via email to