Re: [Roll] WGLC for draft-thubert-6man-flow-label-for-rpl-03

[Philip Levis <[e-mail]>]

On Aug 11, 2014, at 1:26 PM, Brian E Carpenter <[e-mail]> wrote:

> Therefore, treating RPL as a special case is the remaining option.
> But does the ROLL community actually have consensus that they want
> this special case?

I'd argue that if you look at the 15+ so years of work into mesh routing on low power and lossy networks, data path validation is a common technique to almost every approach. That is, there is absolutely consensus in the people engineering these protocols that data packets need to carry information about the route so the network can quickly discover and adapt to changes in the topology. The reason is that the topology can change very quickly, on the order of a few tens to hundreds of packets. 

It turns out that this is very difficult to do efficiently in the IP architecture if you have small packets. HbH headers aren't tiny, and IP-in-IP encapsulation is significant.

In the broader scope of running IP over LLNs, I think that the ability to encode data path information into IPv6 headers without introducing significant overhead is critical. I therefore think this special case is worth it and important enough. But I also think it's important to very explicitly state that it is an additional special case, e.g., through an Updates:.

Phil