Chairs: Carsten Bormann, Ari Keränen
Note takers: Stuart Cheshire, chairs

16:10 - 18:10 Tuesday 26th March 2019

(Meeting started late due to technical difficulties)

Intro, RG Status, Report from WISHI, Pre-IETF work meeting, and Hackathon (Chairs)

Slides: https://github.com/t2trg/2019-ietf104/raw/master/slides/T2TRG-2019-IETF104-slides.pdf

T2TRG has continued with WISHI bi/tri-weekly meetings. Organized coordination meeting with OMA SpecWorks and work meeting on Friday before IETF meeting. Going forward, will continue WISHI calls and coordination meetings with OCF and OMA SpecWorks. W3C WoT workshop June 3rd to 5th. In Montreal planning to have WISHI hackathon and joint meeting with OMA. Interest to co-locate T2TRG event with academic conferences/events in 2019; recommend contacting the chairs if anyone aware of potential candidates.

“State of the Art and Challenges for the IoT Security” draft is in the RFC editor queue. A number of potential new drafts being discussed on edge, CoRE apps, CoRE interfaces, layer 3 considerations, notes from the WISHI work, CoRAL documents. For “RESTful Design for IoT” draft, planning to add new terminology. Also looking for more practical experiences from building RESTful based IoT systems and outside reviews. Two additions to terminology planned: “transfer protocol” and “transfer layer”, terms we have been using for a decade in CoRE WG and T2TRG but not formally defined before. CoAP, HTTP, but also AMQP, MQTT, and XMPP have common features; re-usable and often replacable parts in the stack that are used to transfer data items. Recommendations / feedback on how to define these potential terms very welcome.

In WISHI calls, multiple topics: new object models, IoT extensions to schema.org, re-usability best practices, semantics to data, parsing & translating binary data to make data handling simpler, structured media types for hypermedia exchanges. 5th WISHI hackathon over the weekend with lots of work around CoRAL. CoRAL seems to work well for resource discovery and beyond. We are getting better idea on how all this should work together.

In the pre-IETF Friday meeting, discussed data model component reusability, using CoRAL for hypermedia in IoT, common metamodel language,RG deliverables on IoT security, edge & IoT, and Layer 3 considerations.

W3C Web of Things WG/IG update (Matthias Kovatsch)

Slides: https://github.com/t2trg/2019-ietf104/raw/master/slides/2019-03_IETF104_W3C-WoT-Update.pptx

Update on collaboration with W3C WoT. Looking at standards to complement what is out there and help with interop. Key part: Thing Description. Enables to learn what a thing can do and how to interact with it. Close to completing the work on first set of building blocks. Architecture and Thing Description document going for W3C standards. Scripting API turned to WG note; will continue work in the interest group and get more consensus on this.

Transition to candidate recommendation in two weeks, April 9th. Providing a test suite that shows the document is complete and ready. Then by W3C process to go for proposed recommendation, if successful, 21st May. Then good to tell your W3C AC rep to support publishing this. Standard hopefully ready end of June.

W3C second workshop coming in June 3-5. Should submit “notion of interest” to join or a position paper to get stage time. Deadline April 15th.

Resource Directory Replication: draft-amsuess-core-rd-replication (Christian Amsüss)

Slides: https://github.com/t2trg/2019-ietf104/raw/master/slides/resource-directory-replication.pdf

Exploring distributed Resource Directory setup beyond the simple shared database approach. Tolerant for single nodes to fail but allows for lookups.

Second document is on how we describe which device another device talks to; for example when mobile phone in local network talks to an IoT device and then moves to another network where the address changes and phone’s URLs don’t work anymore. Document “rdlink”, a proposal on this. Builds on RDs conceptually but distributed system. Using “coap+at” scheme and something that is property of the device, e.g., crypto ID, and used to express resource on device so looks like CoAP URL. Could be looked up from DHT. Prior art around HIP and Tor; could learn and take lessons from. Lots of this already in the design here.

Looking for feedback from the RG who else would participate in this. Goal: run with off-the-shelf IoT devices in 2023.

Like to gather general feedback on mistakes made, things not consiedered, and is this RG the right place? And looking for requirements for similar apps.

Bilhanan Silverajan, Tampere University: on first part, RD replication, work that should have been started a long ago. We need that replication in many places and use cases. Seen this being applied in many industries. We see weird implementations today, RD is CoAP server etc. Hope to see this goes forward. The other part, interesting, keen on the protocol negotiation part. Would love to discuss more off-line. Like the idea and prior art is there. Could also look outside of the prior art; OCF has similar schema. Would like to talk about this later.

Erik Nordmark, Zededa: in case of global RD, what are thoughts on level of control who sees what? May only want to expose to my company, family, etc.

Christian Amsüss (CA): Three parts to solution. RD would get registrations accordingly. What do I know from hashes? The Tor v3 IDs are in the hash table encrypted with endpoint name and only one with that given name can lookup anything on that. The authority component. Third part restricting control with respect to organizations; this doesn’t expose devices from the network. Unless there’s proxying, not reachable. Helper nodes do not need to rely on existing network. In border of net could run a node with well-known address. If nodes should not be announced outside, could do enforcement at that point.

Erik Nordmark, Zededa: most you can do is count how many people are sitting behind IP address?
CA: yes. Did look into hiding position when running more advanced helper node. Could do even more but a next next step.

Erik Nordmark, Zededa: could encrypt info in the DHT node if have key

CA: could have helper hide everything

Ari Keränen: seems like RG topic and there’s interest in the group; worth exploring deeper

Carsten Bormann: hallway discussion that comes again; we’re putting IoT devices into jails. Need ways to plumb between these jails. Directory infra might help doing that. Goes beyond having IP address of something; need more machinery to start to talk to node.

Christian Amsüss: not sure of all that but hope we can follow up on this

W3C Community Group on Schema extensions for IoT; schema.org update (Michael Koster)

Slides: https://github.com/t2trg/2019-ietf104/raw/master/slides/iotschema-update-ietf104.pptx

New community group at W3C. Planning to contribute to schema.org based on the events, actions, properties, dataitem, affordances. Plan to drive discussion in Github and community group. When know more detailed what want to contribute, work with schema.org.

One data model is a bunch of SDOs and vendors together creating high level semantic model that doesn’t require specific protocol implementation. Probably converging on the same patterns. See proposals in the Github repository (link in slides). Facilitating eventual convergence on a single model everyone could use.

At schema.org, building the ontology and getting broad agreement on it. Now creating examples with the tool and format of the existing ecosystems. Need to make easy for developers: tools and scripts to automate lots of this. Building web interface for navigation of the schemas.

Want to work with automotive folks and building folks doing schemas today. External Feature of Interest ontologies.

Carsten Bormann: anything you want to report from the OneDM meeting?

Michael Koster: at the first f2f meeting of the group now. Hashing out high level details how to go about with this. Started as low-level thing on picking up a model. Now realizing we need to do something like capabilities, events, actions. Working on examples now. Had a good day yesterday agreeing on how core models should look like. How to compose things. Discussing licenses of different SDOs. Afternoon building representative samples. Using JSON-LD with name spaces for common format. No one else has proposed alternatives so may go forward without lots of arguing how the model should look like.

Eve Schooler: curious to know, who is not in the meeting? In addition to companies are there other orgs in IoT space that you need to bring in and endorse.

Michael Koster: Yes, Apple.

Michael Koster: probably want to enable domain experts to make these. Work with IIC for example. Working on liaison agreement. Want to bring people form IETF/IRTF. Some peope already involved. Also bring people from other orgs, including Apple.

Eve Schooler: would be good to see full list of participants. Exciting activity. Looking forward to the minutes of the meeting.

Michael Koster: we have “invited experts” category if interested to join. Let me know if you’d like to participate and can get invitations.

Using CoRAL to bring more hypermedia to IoT (Klaus Hartke)

Slides: https://github.com/t2trg/2019-ietf104/raw/master/slides/coral.pdf

Carsten Bormann: we should provide a link to the IETF103 tutorial recorded in youtube (https://youtu.be/Uz3CRNjW-NU?t=5070) if you want to have more details on how CoRAL works. Now looking into larger world of how you could build apps in RESTful hypermedia oriented way. And what place CoRAL has there.

Klaus Hartke: will provide some more details on what we did at the hackathon. But first briefly talking about hypermedia at IRTF/IETF. The idea has been around for a while. Many formats based on JSON (usually) proposed. None of them has made it to a WG document. But number is increasing and shows potential here. And shows there’s need for standardization. In CoRE WG the link format showing age; thinking of replacing it with CoRAL. Today primary use case is .well-known/core. Mark Nottingham has related doc at HTTPbis group. CoRAL has two serializations: CBOR based very suitable for constrained IoT – main difference to other hypermedia formats. Also text format, mainly for examples.

Thing description is much longer than CoRAL but also does more than CoRAL; grouping and protocol agnostic parts, and protocol binding. If you are using CoAP, you can reduce the amount of information nicely. Also took full IPSO re-usable resources list and expressed as CoRAL vocabulary.

Conclusion from the hackathon: it works. Next steps to do more experimentation to see where else it could be applied. Earlier discussed group discovery in RD. Also extending link format with new attributes can be elegantly solved with CoRAL. More experimentation is useful – if group has use cases would be interesting to hear if CoRAL can provide solutions there. Also we have bunch of specs: CoRAL and CIRI specs and show cases. CoRAL spec itself quickly stabilizing. Good time to look for a home in a WG.

Carsten: collecting links to other tools using CoRAL?
Klaus: no slide on this, but in the Github repo some links to implementations, e.g. from the hackathon. Github repo also has examples, ABNF, CDDL, etc.

Carsten: side meeting tomorrow 15:00-17:00 in Tyrolka

Ari: personal experience with CoRAL was that it took a while to get my head around it but the more I understand it the more useful it seems. Can highly recommend having a closer look at this and we will post the tutorial material for easier intro to this. Seems to have lots of potential to explore this further.

Meeting planning, Wrap-up (chairs)

Ari: In these summary meetings mainly reporting what has been done. If interested to work on the semantics and hypermedia topics recommend joining the WISHI calls, if interested in the collaboration with other orgs join the collaboration calls. All info on the upcoming events will be posted on the mailing list. Very interested to have colocated event with academia this or next year. If you have ideas on good venues please talk to the chairs.