Re: [6tsch] What is missing slides -- input needed

["Pascal Thubert (pthubert)" <[e-mail]>]

Hello Xavi:

You can expect that people before you (Thomas for his draft and Maria Rita for 1.a) will have described the value of deterministic networking, the capabilities (and limits) of deterministic wireless, and the applications that will benefit from IPv6 over TSCH. So you'll be facing a crowd that already has a sense of a huge value for the real world, even for everyday life.

Your item 5 is the core of the question you are treating. I think that should be most of your time. At the end your 15 minutes, people must also be convinced that there is work to be done, understand what the scope of that work is, and that it is achievable within IETF classical operations. Your list for item 5 is perfectly in line with my thinking. I'd just add security and in particular commissioning and join process. You may want to use the architecture blocks or a subnet diagram as illustrations.

People should have gotten your item 1 in the welcome message from the chairs. As an intro, I'd insist that we aim at enabling new applications such as automation and metering over a converged IP network, for the well-known benefits that we found converging voice and video.
For item 2: It seems to me that the following sub-items are not controversial and position the missing links:
- existing standards found in smartgrid for metering use IETF building blocks but are not designed to operate on TSCH => missing deterministic aspect
- existing TSCH-based standards found in industrial are not compatible with one another => double OPEX, customer unsatisfaction
- they have their own versions of protocols and elements that the IETF defines (PCE/P, ICMP, DHCP, PANA, DTLS ...) => missing the benefits of sharing those building blocks
- but do not support other required functionalities that are available with IETF protocols (RPL, ND, CoAP..) => missing distributed routing
All in all, the lack of IETF building block hinders the network convergence that we are after.
I'd place your item 4 before your item 3, In there, you probably want to indicate that each of the existing solutions address a subset of the problem but that we want a single converged solution, in particular with a mix of centralized and distributed path computation, and a mix of track forwarding and routing.

At the end of item 4 (now 3), the audience would have a nice view of the landscape of existing -deployed- work as a side effect. This can be used to justify the bottom line of your item 3, clarifying that compared to what was already achieved and validated, what we are talking about is mostly the cement as opposed to the building blocks. This gives you an excellent intro for item 5, and you should be there around ETA = 5 minutes.

What do you think?

Notes:

In the thread we discussed that we do not want to use the term proprietary, so as to avoid an inappropriate tension and jitter.

Existing protocols, the fact that the concept validation and a large chunk of the technical engineering were done already, will be revisited in other sections, in particular 2a and 2b. So you can make a forward reference. So please save your time on those.

[cid:image001.png@01CE82E2.970E9C60] [cid:image004.png@01CE82E2.36DE4840]


Cheers,

Pascal

From: Xavier Vilajosana Guillen [mailto:[e-mail]]
Sent: mercredi 17 juillet 2013 02:04
To: [e-mail]
Cc: Kris Pister; Thomas Watteyne; Pascal Thubert (pthubert)
Subject: Re: [6tsch] What is missing slides -- input needed

Hi Kris, Pascal, all,

some of the ideas drawn on this discussion are confusing me, sorry!. I like to be very synthetic and write the main points as I have on my mind:
1-we aim to define how to run IPv6 on IEEE 802.15.4e TSCH.
2-An analogy of our work can be found on ISA100.11a and WiHART approaches although the flavor is different as they don't use IETF building blocks. In fact these blocks have not been though for IEEE 802.15.4e, so while people may think the work is done, that is not case -- and this goes against our work as we need to convince as well--
3-For our case, the blocks we need to use already exist (zigbee IP, etc...) however the guidelines on how/what exactly glue together are missing.
4-The latest made vendors to implement their own variations of these blocks that limit open-inter-operability or at least make it harder.
5-Our aim therefore is glue all together and glue it with the IP architecture including what is on the backbone (ND, etc..)
    The main points that need to be glued together are:
         -Scheduling construction and maintenance:
                     -Centralized vs distributed: support for both
                     -Protocols to distribute the information: Based on PCE architecture or distributed (RSVP like)
          -QoS and TE due to the cell nature of the underlying MAC layer (including priorities, Traffic Classes, etc..)
          -RPL on TSCH, routes and best effort vs deterministic tracks: support to mobility and fast commissioning
          -Security, key distribution, EB security
          -ND and backbone architecture
          -Packet switching and tunneling
          -management and maintenance of TSCH, seen as an orchestra director based on underlying TSCH information (6top)


do I miss any points? :-)

thanks!
Xavi


are these points covering the topic? What else I'm missing.

On Tue, Jul 16, 2013 at 2:56 AM, Pascal Thubert (pthubert) <[e-mail]<mailto:[e-mail]>> wrote:
Dear Kris and Xavi:

Please see below.

I agree with you on all of this.
[] Same here; I think that Xavi's text is well-balanced. And we will not have time for controversies at the BoF so this discussion on avoiding them is critical.
We're trying to make a fully IETF/IEEE compliant standard, and there are definitely some missing blocks.  We can point to some industrial standards that have some similarities, to say "see, it can work really well, but they don't do us any good because they aren't based on IEEE 15.4e".
Our goal is to tell people how to run IPv6 on 15.4e.
[] Yes, that is the primary goal. Hopefully the architecture that we are producing applies to all sorts of TDM-based LLNs. The key here is to leverage the new level of guarantees and services that Time Slotting enables.

I just have a pet peeve about the IETF calling anything that isn't IETF "proprietary".
[] It's true that IETF has its view of open standards that is useful for discussion within the IETF, but 1) your words are not true to that definition (see section 7 of RFC 2026) and 2) it is not just the IETF that redefines the term for its own use; it's really all sorts of countries and organizations, often referring to the way the standard was elaborated and is made available, in particular with regards to fee policies, IPR rules, or open source implementations. Still I agree that this time and place are not for fighting on the definition of such terms. If the IETF term is not the right tool for our discussion then let us be more specific on why we can't just live with a status quo.
The existing protocols:
- are not fully compatible with IEEE 15.4e which came later
- are not compatible at all at any layer though similar on paper
- are not implemented as open source
- do not define all the required interfaces and protocols for operation over the Internet (e.g. interface to the WiHART AP or ISA100.11a backbone router operations over the backbone)
- have their own versions of protocols and elements that the IETF defines (PCE/P, ICMP, DHCP, PANA, DTLS ...) and for which open source implementations are available
- do not support other functionalities that are available with IETF protocols (RPL, ND, CoAP..).
- ... ?
Just like proprietary RPL implementations over TSCH, industrial standards demonstrate both the need and the feasibility of our work and show the way.

And I don't think that we should try to point to WiHART and ISA100 and say "they are missing some blocks" because they are outside of our scope anyway.
[] Agreed. At the same time, we need to explain why we're not all set with what exists already on the market. We must make sure that we do not make it look like criticizing the excellent work that is really our foundation. On the contrary we want to strongly express that reuse all the concepts that those protocols have demonstrated but will base the devices interactions on existing IETF (open ; ) standards when possible.
Makes sense?
Pascal
On 7/15/2013 5:29 PM, Xavier Vilajosana Guillen wrote:
thanks Kris, this makes sense. Last Friday on the phone call we kept semi-propietary "word" because there are some vendor specific blocks that although are defined do not match between different  vendors (e.g way schedule is distributed, at least in 15.4e). I think it is already defined in WHart and ISA100.
The points you raise here are very important as this is what some of the people that will be listening at the BoF will raise. So we have to be very sure and be very careful on how we present this.  The idea of that presentation is to outline what is missing and why we go for 6TSCH as a "glue" for all the blocks.

so my comments inline:

1) they are based on TSCH, but not IEEE 802.15.4e TSCH.  4e did not adopt either WiHART or ISA100, rather we/it came up with its own way of doing things.
agree.. I clarify that. I also clarify that we go for defining how this blocks are built on top of IEEE 802.15.4e. (right?)

2) I don't think that even the IETF gets to say that either of these protocols is semi-proprietary anymore.  they are both IEC standards, and they are interoperable across many vendors (of the wireless part).
In the case of IEEE 802.15.4e networks,  what about centralized scheduling distribution or distributed scheduling. What about how security is installed at each node, what  about QoS maintenance (including overprovisioning, or cell reallocation). Maybe vendors inter-operate ( in ISA100 and WHART, not so sure in 15.4e), but are aiming to propose a common approach for that right?-- this is because the existing approaches (in the case of 15.4e) are vendor specific ... I guess this is what we want to show in the BoF, everything exists but we aim to find a common direction for everyone.

3) all of the blocks are defined, and there aren't any missing.
These are complete standards-based solutions, they just aren't based (completely) on IETF/IEEE standards.
well because some parts are not defined by IETF yet right? e.g messages on the air to schedule one link with a neighbour.
X


On Mon, Jul 15, 2013 at 4:55 PM, Kris Pister <[e-mail]<mailto:[e-mail]>> wrote:
Xavi -
 regarding the existing industrial implementations of TSCH, I'd say things a little differently.
1) they are based on TSCH, but not IEEE 802.15.4e TSCH.  4e did not adopt either WiHART or ISA100, rather we/it came up with its own way of doing things.
2) I don't think that even the IETF gets to say that either of these protocols is semi-proprietary anymore.  they are both IEC standards, and they are interoperable across many vendors (of the wireless part).
3) all of the blocks are defined, and there aren't any missing.
These are complete standards-based solutions, they just aren't based (completely) on IETF/IEEE standards.

ksjp

On 7/15/2013 4:03 PM, Xavier Vilajosana Guillen wrote:
Dear all,

I am working on what is missing? presentation slides. I want to go deep on the description of the following points. From what we listed last webex I developed the content but I need further input. Please update the following points with your thoughts: (all ideas are welcome, I will filter later!)


Deterministic wireless over TSCH is demonstrated and available but semi-proprietary (TSMP, ISA100.11a, WiHART)
   - Vendors have semi proprietary solutions as there are some blocks missing:
       -These blocks are mainly on the upper Data Link Layer and its integration with the network layer.
       -These blocks are mainly for the management and operation of the network
       -missing blocks limit interoperability and mass scale adoption of the technology
       -...
(give me some feedback here please)


Same for RPL/TSCH with scalability to *1000s
   -Vendor specific RPL proved to scale to 1000s (need info about that please!)
   -TSCH networks proved to scale to 1000s
   -Missing junction between both. RPL on TSCH.
   -....

(give me some feedback here please)

Most IETF components exist (ZigbeeIP)
    - ZigBee IP Supports 6LoWPAN for header compression, IPv6, PANA for authentication, RPL for routing, and TLS and EAP-TLS for security, TCP and UDP transport protocols.
     -the building blocks exist but need to be fit to TSCH nature. Slotted and deterministic MAC layer, mapping of RPL routes to TSCH schedules (that globally build tracks.) -- Coexistence of Tracks and routes.

(give me some feedback here please)


Missing IETF architecture to put it all together
   - Pascal Picture from architecture draft.
   - Coexistence==Support of PCE/ Distributed in same architecture

(give me some feedback here please)
Thanks!
Xavi


_______________________________________________

6tsch mailing list

[e-mail]<mailto:[e-mail]>

https://www.ietf.org/mailman/listinfo/6tsch


_______________________________________________
6tsch mailing list
[e-mail]<mailto:[e-mail]>
https://www.ietf.org/mailman/listinfo/6tsch