Thing description task force -- 11 Nov 2015

<michael> So far so good!

<michael> webex not starting?

<dsr> scribenick: dsr

<inserted> @@@Darko's slides@@@

Darko: I would like to recap where we got to at the Sapporo face to face.

<kaz> [just in case: Kaz, Dave, Darko, aparna, Darko, Frank, Katsuyoshi, Kotakagi, Ryuichi, Taki, Toru, Victor, Yingying, Michael, Arne, Danh, Louay]

He presents a summary slide

<kaz> f2f minutes (still need to be tidied up)

Darko summarises the plug-fest. Some 8 or 9 implementations. Focus on using thing description as a basis for interoperability

These descriptions were made available on GitHub

The use case centered on an RGB lamp

We had some 14 different thing description files. Almost all using REST. For next time we should have some none REST based approaches too,

Some discussion on additional semantics and on hypermedia controls. We could discuss this in future calls.

Let us know if you want to raise this.

For the next plug-fest, we want to hear your ideas.

kaz: do you want to mention the IRTF T2TRG interest in participating in a joint plug-fest exercise

Darko: yes, they would like us to organise a joint meeting for that.

<kaz> IRTF joint meeting minutes (still need to be cleaned up)

T2TRG are working on HATEOAS and approaches to representing resources

We discussed that the HATEOAS approach should be supplemented with a thing description

<inserted> scribenick: kaz

dsr: had a meeting with OIC open source group about collaboration

darko: would you invite them to the TD call? or the whole IG call?

dsr: they have some resource description
... and binding for protocols
... that's one thing to consider
... there are interested in following the TF work
... might come with some representation

darko: another point from the Yokohama joint meeting
... prepare for the next plugfest during our IG calls
... mapping, thing description, etc.
... inviting them to our calls would be useful for the preparation for the next plugfest

<inserted> scribenick: dsr

Dave: we should consider what it would take to transform between different representations — what problems emerge - what is present or absent in each representation

Darko summarises the break out session. The first part covered the feedback from the plugfest, and the second covered the relevance of EXI for efficient message encoding

<DarkoAnicic> http://www.w3.org/WoT/IG/wiki/Breakout_topics_30-10-2015

We also heard some results for data compaction using different approaches

These were raw JSON, CBOR and EXI4JSON. The latter compresses raw JSON to a quarter of the size.

(for the test messages)

<kaz> dsr: it's question of computation as well, so could be 3-dimensional score

Dave: we should try to compare the encodings on three dimensions: message size, RAM and CPU resource requirements.

Daniel: Yes, that would be desirable for future work.

Darko: how could that be done?

Daniel: in the EXI WG we set up a test framework and measure the time needed to process the test files.

Dave: I agree, measuring the RAM usage is a little harder but doable.

Darko: how was the conversion from JSON to EXI4JSON done?

Daniel: there are techniques for mapping JSON to XML.

Darko: why didn’t you go JSON-LD to XML-RDF and then to EXI?

Dave: I have done some experiments myself on efficient encodings and would be happy to report on that in future

Darko: okay when you are ready, we can put it on the agenda
... any other comments/questions? [no]

Darko switches the topic to the part of the break out on extending thing descriptions with additional context

We talked about additional domain dependent and domain independent context info

<kaz> brightnessProximitySensor.jsonld

An example is “who are you”. A name is insufficient unless you relate it to an ontology with further background

Darko presents a layer cake diagram for contextual semantic models

Dave: provided we have a way to map terms in the JSON to Linked data URLs we have plenty of flexibility

Darko: some people looking at thing description examples have missed this point

Michael: how do we use semantic models that describe things via ontologies

schema.org is a web resource with vocabularies covering range and domain.

It is a little looser than OWL ontologies

Darko: schema.org is widely used and we could look at how to exploit it

We should elaborate some examples as a basis to explain what is possible

This could be a goal for future plugfests

Michael: ontologies allow us to have scalable descriptions

<inserted> scribenick: kaz

darko: during the encoding process, useful to handle mapping with existing model
... what I mean here is application model is something specific to apps
... some particular application domain like smart home
... maybe that layer is not in scope of our (=WoT IG) standardization

victor: in the api tf, people are also talking about that
... so kind of strange to hear application model layer is out of scope
... maybe a bit different name should be assigned to avoid confusion

dsr: vertical domain meta data?

<dsr> scribenick: dsr

Michael: there are existing organisations that are working on domain specific vocabularies

Darko: we should also look at other approaches e.g. HATEOAS, Hydra, RESTdesc, RAML, etc.

Dave: can we clarify how these compare to the role of RDF as a general metadata framework?

Darko: our current thing description examples lack details on which protocols can be used with specific properties

<kaz> s/Daniel: how was/Victor: how was/

These other approaches are at a lower level and offer such specificity for protocol usage

Michael expands on the kinds of metadata involved

… the notion of APIs goes away

Dave: do you mean we should give applications direct access to lower layers in the abstraction stack?

Michael: yes

Dave: the ability to enable apps to be decoupled from the transfer and transport layers is made practical through rich communications metadata that servers can exploit to suit the context

Darko runs through the mapping from thing descriptions to REST

The choice of REST methods can be determined based upon the thing description

Michael: given that we need to embrace a wide range of existing systems, then giving the server the means to adapt has value

<DarkoAnicic> http://www.w3.org/wiki/TPAC/2015/SessionIdeas#Space.2C_Time_and_Sensor_Semantics_in_the_Web_of_Things

Darko switches to the cross group meetings that took place in Sapporo, e.g. with the automotive and spatial data groups

Kerry Taylor and colleagues talked to us about space, time, and sensor semantics. The spatial data on the web WG plan to standardise the semantic sensor network ontology

<kaz> IRC minutes (to be cleaned up)

Darko: the last thing on my agenda is the tech landscape which we need to start work on
... I have started a wiki page as a collector. This has a table with columns for various criteria

<kaz> TF-TD Tech Landscape

We could have several such tables

Dave: I would recommend that we have a separate summary for each approach. We could check these with the originating organization. The tables are great but too constraining to be the whole thing.

Michael: I agree

Darko: so you will do that the IPSO, right?

Michael: once I have joined the group, so I get write access to the wiki, right.

Darko: any further comments/suggestions? [no]

Dave: we want to include people’s names as a reward and to enable peer pressure on finishing each task

Darko: we need volunteers …

Dave: one suggestion is to switch to GitHub and use markdown, so that people can use pull requests and we can track issues more easily

I would do this once we have clarified that document structure

Kaz: during the face to face we have several volunteers

<DanhLePhuoc> count me in for Thing description

Dave: suggests doing a poll to see who knows about which approaches as a stepping stone to signing people up as volunteers.

<DanhLePhuoc> and also 2.Metadata model

Darko: we will discuss and track the tech landscape study in this call

Dan: if we have the volunteer’s names on the wiki that would help

Darko: okay, but we need further volunteers

… end of meeting …

- DRAFT -

Thing description task force

11 Nov 2015

Attendees

Contents

Summary of Action Items

Scribe.perl diagnostic output