Re: [6tisch] Comment to draft-wang-6tisch-6top-coapie-01.txt

[Tero Kivinen <[e-mail]>]

Thomas Watteyne writes:
> I'm cautious in embracing a standard without understanding benefits and
> overhead:
> 
> Where can I download the 802.15.9 spec? All I can find is 
> https://www.ietf.org/proceedings/84/slides/slides-84-roll-3.pdf

The 802.15.9 is now in the Sponsor Ballot, and we are not allowed to
distribute it to others now, but you can go and buy it from the IEEE
Standards Store:

http://www.techstreet.com/ieee/products/1902465

The price is $86, which is not that bad, compared that the
802.15.4-REVc draft is $433
(http://www.techstreet.com/ieee/products/1894072).

> What is the scope of 802.15.9?

The 802.15.9 related documents can be found from mentor:

https://mentor.ieee.org/802.15/documents?is_group=0009

and the scope of 802.15.9 can be found from the latest PAR change
document:

https://mentor.ieee.org/802.15/dcn/15/15-15-0464-00-0009-p802-15-9-par-detail-draft-change-2015-05.pdf

I.e the scope is:

	This Recommended Practice defines a message exchange framework
	based on Information Elements as a transport method for key
	management protocol (KMP) datagrams and guidelines for the use
	of some existing KMPs with the IEEE Std 802.15.4. This
	Recommended Practice does not create a new KMP.

> It is presented as a Key Management solution, yet you suggest to use
> 802.15.9's mechanisms for (i) IE space and (ii) fragmentation.

The 802.15.9 document is split in two pieces, the multiplexing and
fragmentation layer, and the key management layer. The key management
layer requires a method of transmitting messages which are larger than
802.15.4 phy frames, thus it needed layer to fragment KMP messages and
transmit them over 802.15.4 phy frames. While we created that, people
said that this mechanism would be useful also for others, so make sure
other protocols can also use this. Because of this we added simple
multiplexing method, which will allow others to use the fragmentation
and transport method too. 

> It looks to me like we would be exploiting a tiny fraction of a
> much more involved standard, but have to include the full memory footprint.

Actually most of the 802.15.9 implementation parts are in the
multiplexing and fragmentation layer. The KMP layer is very minimal,
only consisting of one byte identifying the KMP protocol to use. I.e.
MP IE Format description in the specification is 4 pages, and the KMP
service description is one page...

All the actual key management protocols are described in the appendix,
and none of those are mandatory to implement (i.e. all are informative
annexes). I.e. you can implement 802.15.9 without implementing any key
management protocols, and if you do not have any KMPs configured, then
all your MLME calls related to them are not really relevant. 

> How much flash/RAM memory is a 802.15.9 implementation?

As this is recommended practice that is still in the draft phase, I do
not think there is any implementations of it out there yet. There are
other groups interested in using it, so perhaps they will get some
numbers out after we get through sponsor ballot.
-- 
[e-mail]