Re: [Roll] Working Group Last Call: draft-ietf-roll-protocols-survey-02

On Dec 5, 2008, at 8:49 AM, Ian Chakeres wrote:

> I've reviewed draft-ietf-roll-protocols-survery. I have a few  
> comments and nits.
>
> In the Introduction, it would be great if LLN constraints and
> requirements that other networks do not poses were listed/summarized
> before reference to the routing-req docs. The existing text left me
> guessing about what exactly was different.
>
> I would suggest rephrasing the DYMO discussion related to hopcount in
> Section 3. Perhaps stating that "distance" is the metric in DYMO, and
> that hopcount is the minimum distance allowed.
>
> In Section 4.2, the motivation provided is the possibly large network
> size (250-millions) and deep network (20 hops). Network size and
> routing information size do not have a direct connection when
> hierarchical addressing is used. Perhaps this motivation could be
> removed and simply the last sentence kept. Given this change maybe the
> suitability metric should become routing table size, instead of table
> scalability.
>
> Based on the suitability table and description contained in the ROLL
> survey document, I have updated the DYMO (v16) draft .  First, I added
> text to indicate that distance may be influenced at multiple points in
> the processing of RM. This changes should allow link and node metrics
> to influence distance before and after RM processing. Secondly, I have
> made forwarding of RERR optional (through strongly suggested). This
> should allow a smart implementation (e.g. using an "active" precursor
> list) to limit the propagation of RERR messages. Thank you for
> pointing out these minor issues in DYMO.
>
> If you have any other suggestions for the DYMO specification that you
> feel would make it better fit to suit ROLL requirements, please let me
> know.

Ian,

I'm glad our reading of the draft was helpful. We'll be sure to update  
the text to reflect your edits and points.

We've gone back and forth a lot on DYMO's relaxation of link cost  
computation, so it's great to hear DYMO's author's take. :)

Your point about hierarchical addressing is interesting, and I think  
it relates to comments others have made on fisheye routing and other  
state suppression approaches. Help me out if I'm mistaken here, but I  
don't recall any work on address allocation/distribution in wireless  
networks to enable good route aggregation. This is particularly  
difficult in the IPv6 case because of how addresses are specified. So  
I'm wary of assuming hierarchical addressing, when in some cases it  
might not be possible. It's not quite like the wired Internet where  
there is some aggregation due to the administrative domain nature of  
how addresses are allocated.

Phil
_______________________________________________
Roll mailing list
Roll@ietf.org
https://www.ietf.org/mailman/listinfo/roll