Minutes Webex 10 October 2013, 6TiSCH WG, models draft team
Note: timestamps in PDT.
Taking notes (using Etherpad)
- Xavi Vilajosana
- Thomas Watteyne
- Maria Rita Parattella
- Pouria Zand
- Qin Wang
- Raghuram Sudhaakar
- Rouhollah Nabati
- Thomas Watteyne
- Xavi Viljosana
- [08.04] meeting starts.
- [Xavi] Discussed with Raghuram the idea of adding section 3 called "generic data model". The idea is to fill that with YANG for data model for this version of the draft. After the Vancouver meeting, this section can be moved to a (future) draft dedicated to the data model. The idea is not to have the DM be specific for CoAP.
- [Xavi] By now draft will focus on CoAP DM. After Vancouver, we can split and have YANG DM, Will generalize later.
- [Thomas] Is the DM currently following 6top closely?
- [Raghuram] The DM contains a set of exchanges. In YANG, we will define hierarchical definitions of the different fields in every message. IM will exchange transaction. For CoAP, there was a general consensus on the ML for CBOR. All of these messages will have a CBOR representation. A "GET URI" might not contain any payload, since the expected action is fully defined by the GET method and associated target URI. Generic model in YANG; includes a tree-like definition for request and response.
- [Thomas] It would be good if the DM could be used for other communication paradigms, such as 6top.
- [Qin] We identified several communication paradigms identified (L2, L3, L5). Can you refine?
- [Thomas] The question is whether it is possible to use the generic model to describe the content of the 6top-to-6top interaction, too. If that's the case, the same (YANG-based) DM could be translated to different communication paradigms.
- [Qin] 6top-to-6top communication uses IEEE802.15.4e information elements. Is that not a problem?
- [Thomas] One could things of the IE as a container for payload.
- [Xavi] 6top defines the IEs itself. We could have the content of those IEs defined by same generic DM.
- [Thomas] Extend that to RSVP-like communication paradigm, others. 6top can be seen as just a set of functions that one can call. The data model is generic and can be translated to different "transport" mechanism.
- [Raghuram] Do we want to have the same data model for coap and 6top?
- [Thomas] What we could do is, for now, define it for CoAP (since we have a time contraint). Later, we can take this data model out and use the data model for any transport including L2, L3 and L5.
- [Raghuram] The DM in the current draft is closely tight to CoAP in many ways. Not only as a translation but also functionality aggregation.
- [Thomas] Extensions could live in 6top as well.
- [Raghuram] Moving on to 2nd part of the conversation: extensible resources. We want to define methods for users to define events. That involve a set of 6top operations. Methods to express what the user wants to monitor. Augmented Binary Normal Form (ABNF) looks like the best option. Clear grammar to express rules. Simple equations. Logical operations.
- [Xavi] we start with CoAP. When we split with generic DM, this can explain with composition. Rules for composing those messages to request multiple events.
- [Raghuram] slightly different. We don't need to describe any rules, just some example. ABNF will allow this.
- [Xavi] Agreed, but from the implementor side, just need to define regular expressions.
- [Raghuram] ABNF could play the role of the "regular expression". Subset of functionalities can be defined with profiles.
- [Thomas] How to distinguish complexity? Let's make sure it's not too complex.
- [Raghuram] If it is too complex, a devic not implementing ABNF could be prevented to have extensible resources.
- [Thomas] In this context, what is a profile?
- [Raghuram] A profile is a colleciton of such URIs, which reflect the extension resources.
- [Thomas] YANG model will generalize relational model sent by Pouria.
- [Pouria] Any questions/comments about the data representation sent to the group?
- [Qin] About cells. In the current 6top draft, in a cell we do not provide statistic data.
- [Pouria] How can be include the relation between the tables in the draft?
- [Thomas] Plain english might be a good start. Contains the same information as the 6top draft.
- [Pouria] Use simple tables? or use a YANG data types to express the relations with components.
- [Raghuram] Will push latest version to the repo by tomorrow's call.
- [Raghuram] Pouria, can you do the same with section on data?
- [Pouria] Will do.
- [08.43] meeting ends.