Minutes IETF108: asdf

Meeting Minutes A Semantic Definition Format for Data and Interactions of Things (asdf) WG
Title Minutes IETF108: asdf
State Active
Other versions markdown
Last updated 2020-08-03

Meeting Minutes



TIME: Tue   2020-07-28  11:00UTC

materials:    https://datatracker.ietf.org/meeting/108/session/asdf
agenda:       https://www.ietf.org/proceedings/108/agenda/agenda-108-asdf-00

meetecho URL: https://meetings.conf.meetecho.com/ietf108/?group=asdf
jabber:       xmpp:asdf@jabber.ietf.org?join
mailing list: https://www.ietf.org/mailman/listinfo/asdf

etherpad:     https://codimd.ietf.org/notes-ietf-108-asdf?edit


97 people in jabber room

1) Welcome and Note Well

### Niklas presenting Agenda.

2) hellos and mic check

3) Intro; brief introduction into OneDM, SDF (Proponents);
   (and clarifying questions) [30 minutes]

### Carsten Bormann (CB) presenting.
- We currently have lots of diversity in IoT systems that doesn't provide much
value. - Objective is to harmonize data models. - No need for another wire
format, rather harmonized data and interaction models. - Some components
already in place (SenML, CDDL, W3C TD...) however they solve different
problems. - Avoid creating another standard. (xkcd 927) Ecosystems should not
see ASDF as a competitor. - OneDM recently came out: https://onedm.org ,
started as a liason after ZigBee Hive meeting. SDOs operating under NDA, one
liason per org is complex, came out in order to have a more open process. -
Biggest achievement was legal agreement. Using open source license (BSD3)
allows cooperation on common models without jumping through legal hoops. Second
thing: common specification format SDF v1.0. Now getting data models published
using SDF. SDF not a paper tiger: people have taken SDF and looked into how can
convert models to SDF. Some also doing conversion to other direction. - SDF as
common format for collaboration (the "red star"). SDF is a domain specific
language represented in JSON. SDF defines data models inspired by
json-schema.org. SDF interaction is based on 3 types of affordances: property,
action and event. - OneDM will be one of the users of SDF; others are the
different ecosystems - YANG is not optimized for converting models from and to
it. YANG is focus of people who want to define full-fledged models. Main user
of YANG is the IETF itself. At high level may look similar, but if look how
it's used, it's quite different. - Proposal to standardize starting from
draft-onedm-t2trg-sdf-00; after which gaps in functionality, references,
usability will have to be identified. - The spec currently in many places needs
to be still improved. Gaps in functionality, for example composition. Some
things in SDF being used are not standardized yet; could copy in the doc or get
the normative refs standardized. Work remains.

Cullen Jennings (CJ): could you say a bit more about what those things are?
CB: JSON schema org the most daring. Using that to define the data schemas.
Some others too, but will have to look CJ: close enough, thanks!

- We can also still change SDF. It's not rubber stamping. OneDM is working on
harmonizing things. Possible to work witht he contributing orgs to develop SDF.
Right now things are still wide open. SDF will interact with tools.

Andre Bondi (AB): Question on the inclusion of performance characteristics. It
would be useful to know about performance and communication pattern
requirements. If have centralized way to collect, we can understand what is the
burden in the central controller. How stringent response time requirements.
Other non-functional requirements? CB: interesting idea, maybe not that easy to
capture non-functional requirements. It could be useful if we manage to do it.
AB: useful for capacity planning purpose CB: can be only analyzed with protocol
bindings. Will need to go further there before making useful statements. Agree,
we should examine that.

- Why standardize this now? OneDM has completed a usable doc and happy to
transfer change control to IETF. - Why standardize at all? OneDM contributors
and implementers need a stable, well-defined spec. - Why IETF? Vendor-neutral
process. Ecosystem SDOs already use IETF specs. IETF has some experience with
domain-specific data modeling. - What would ASDF WG do? Focus on delivering SDF
spec in RFC form, ensure that normative refs are stable, work with other

Bron Gondwana (BG): Given the timeline which is of few months, it seems
incompatible with not forming a WG. Which is the timeframe? CB: not something
like January CES but getting started in months would be useful. Getting stable
SDF 1.1 or 2.0 out of WG end of next year would be very important. BG: So it
sounds that it is a WG forming group in all but name. Charles Eckel (CE): how
see translation happening, and models defined in SDOs. If IETF would make
changes. How to make sure translations stay valid. CB: It is the job of OneDM
to actually get the harmonization process defined. We are discussing a document
on how the process should work in detail. Some of your questions will be
answered by that document. Actual tools will be written by ecosystems and
vendors there. Need to keep up relationships there. They will be important
voice on what changes can be made. Need to keep them at the table. CE: Tooling
was indeed very important for YANG and we have also seen it to be very
important for Open Source. CB: some of these tools have been made at IETF/WISHI
hackathons. We have built components that went into those tools.

4) Views of contributing ecosystems (Bluetooth, OCF, OMA [LwM2M], Zigbee) and
   a few interested vendors (…); clarifying questions [20 minutes]

- Alan Soloway (AS) presenting the OMA BoD pov. Overview of the orgs (DM and
IPSO) and Liason Statement signataries that support to the OneDM work. BSD-3
clause license is also already. Presenting a tool that translates from SDF to
LwM2M. - Michael Koster (MK) presenting the pov of Zigbee alliance (BoD) and
Project CHIP (Steering committe). Zigbee Use Cases; need to provide a
friendlier entry point than XML for the Cluster libraries and SDF would be a
good match. - Wouter van der Beek (WB) presenting OCFs pov. BSD-3 clause
license is also already. Created a conversion tool between OCF and SDF. Liason
endorsing the work on public website. - Szymon Slupik (SS) precenting bluetooth
pov (chair of Bluetooth SIG mesh WG). A focus area on smart buildings, which
they believe that it is good match for SDF work. Bluetooth SIG endorses this
work. They are alooking into using SDF as the native environment to represent
data models. - MK presenting uses cases from the SmartThings pov. Objective is
to use SDF for device integration, capability model, 3rd party API integration;
using swagger and WoT for it. - Ari Keranen (AK) from Ericsson pov. Ericsson
uses SDF, see it as a valuable tool to reduce integration costs. Having a
common language that would reduce the cost. Another use is to using SDF
together with LwM2M/IPSO. Already have contributions with tools for model
development and translation.

5) Discussion (beyond clarifying questions) [30 minutes]

Questions to come:
Do we agree on the plan?
Do we have energy to do this?
Should it be done in IETF?

CJ: Regarding change control. How much are we open to significantly change SDF?
Or people want to give it largely as is. Is this starting point or close to
done? CB: This spec is actually and unusually open-to-change specification.
People usually don't directly write SDF but use tools to translate to/from. We
did sweeping changes leading to 1.0 and was relatively painless. Not free to
do, but certainly possible. AS: The answer is actually "both". It is a
well-stablished floor in which we can start evolution. We have a great
foundation but we need to work on it to evolve. Michael Richardson (MR): Kinda
inheritance (?) CJ: didn't argue for that model, just example of a significant
change WB: We have a working situation in which SDF fits multiple ecosystems,
at some point we need to draw a line in order to get started. Is this enough?
No. But this kind of work will never stop.

Andre Bondi (AB): Worked for a long time on performance and found that it is
often a problem, it would be good to embed performance characteristics
information early on. Maybe something needs to be communicated at some point of
time. Seems this could be right place to capture that.

NW: interesting topic. Not right now focus of group but could be extension.
AB: happy to take issue off-line
CB: we have asdf mailing list where could discuss this and have other people

6) Calling the questions [10 minutes]

NW (as chair): was  also asked in Jabber why not WG forming BoF. There was wish
to see more industry support for this. That's why had lined up industry folks
here. This work has been progressing and is reasonably mature. Don't hear any
disagreement -- Barry to comment as AD? Barry Leiba: going well. Let's get
sense whether people think IETF is the right place to do this

(in jabber: a few. Plus FORTISSIMO result)

1.- Draw up a charter and capture what was discussed in the meeting. Any
comments against? No comments. 2.- Do we have energy to do this? User industry
seems to have energy; have expressed strong interest in this. Is there enough
IETF'er to do work on this? Bunch of highly-engaged IETFers in IoT space on

BL: could take a hum for "if you would be participating in this work, hum"

NW: first question: if this work starts, would you participate (use the hum
tool -- to the left of the chat icon)

(result: Fortissimo)

BL: already asked objections, wasn't any. Fortissimo for "i would work on this"
is promising

Eve Schooler (EV): can we have some discussion about where other than the ietf
would this work get done?


CJ: since no objections, this hum may not give that much info
BL: we could discuss what are the alternatices

NW: OneDM folks, as Carsten stated, think IETF is the right place. If we want
to tell SDF folks to go elsewhere there should be objections to not do it here.
If have objections please come to mic line.

{mic line empty on alternatives}

NW: still needs charter etc, but as I see it there's no enthusiasm  to discuss
alternative places to do this.

CJ: has been really well run bof and  convincing. Why not have the charter
discussion here and now?

BL: this came from BoF approval call; perhaps remember there were some concerns
about industry pick-up on this. If there's broad interest. Approved as non-WG
forming BoF with that in mind. Sounds to me it's pretty clear we should send
charte to list, finalize that, and get it to me.

CJ: agree with the plan!

Pete Resnick: seems there's charter in back pocket to work on.  Anything
controversial there? Any particular issues where need input?

CB: maybe we should show the charter (even if against protocol).


PR: no need for details; anything you think need group input? Anything in
particular need help with?

CB: interesting thing in this work -- Last 3 paragraphs. Interesting thing of
SDF: stew of different technologies. Input coming from various places. Some can
be handled separately, like JSON path. Don't know yet if need it beyond
discovery. Things like how we do data modeling and interaction modeling parts.
Hope we can handle all the questions in the WG. Sometimes IESG wants decisions
made before WG is formed so that WG doesn't get blocked by bike shed
discussions but don't see that as problem here. Way we got to SDF 1.0 was very
constructive process. Some WG and RG relationships, like CBOR for CDDL specs,
formal description languages RG, etc.

BL: only thing  that strikes is to  "work with OneDM and contributing orgs"
will get questions  from IESG. Also concerns about there being representatives
that say "OneDM says this" rather than individuals working on this. 
Re-spinning that might be useful cosidering how  IETF works. PRefer to have
people who care to be in the  WG doing the work.

CB: yes, don't want to do this based on liaison statements. Limited usefulness
with LSs. That text should be changed, agree.

NW: Please sign up for the ASDF list (asdf@ietf.org) and for the charter
discussions if you have interest on this. So at this point we have a plan

CB: if you are interested in SDF there are various  tutorial resources you  can
use. One in my youtube channel. There's onedm.org as landing page for info
about OneDM. Also find github repos for various models. All work in progress,
not finished. Needs lots of work but a good way to get familiarized with this.

Georgios Karagiannis (GK): I'd like to know whether we have looked into other
SDOs like OneM2M and W3C. Have looked at what has been done there? CB: Yes.
Have even slide about W3C at the backup materials.

MK: as individuals and also as org have worked with the W3C WoT group. We are
focusing on the domain specific vocabularies that are explicitly excluded from
WoT. Metamodel very much aligned. W3C plugfest has been using OneDM definitions
in thing descriptions. TD applies descriptions to over the wire data formats.

GK: agreement between W3C and OneDM?
MK: yes, currently working on concept called TD templates that can make TDs
composable together with OneDM definitions. Also other intersections like
discovery: the terminology used can be OneDM URIs. Also other vocabulariess
like iotschema can be used. GK: and OneM2M MK: OneM2M was involved early on but
had to partition the work due to Huawei issue. But now not issue. OneM2M is
potentially part of this. Also modeling part of OneM2M and OneDM has another
potential intersection. GK: also agreement here? MK: was early on but has not
been lot of participation. Now being open and everyone can participate we need
to get OneM2M back to participating. CB: T2TRG has been  organizing number of
informal events where we bring people from SDOs and orgs together. Have had
lots of common events with W3C WoT. WoT is bit of sister org of T2TRG. Next
WISHI call day after tomorrow where we talk with MSFT about DTDL. Another
language we want to take with SDF. Will have discussion there. Coordinates in
T2TRG mailing list; can send invite to ASDF list too.


NW: OneDM coming out as visible org, there are many reasons, one was to be able
to address more orgs than in the original liaison. This is about reaching out
to other orgs and incorporating their requirements.

NW: Seems we are done. Any last thoughts?

BL: thanks for the chairs -- very well run BoF!

NW: Thanks!