<scribe> scribenick: kaz

Prev minutes

Feb-24 minutes

McCool: (goes through the previous minutes)
... any objections to publish them publicly?

(none)

McCool: the minutes approved

Semantic discovery

McCool: Daniel, any material?

Daniel: not really...

McCool: we need to look through our history and collect resources
... discussion on directory service and query
... what kind of query should be for discovery
... thought the work was done by Siemens

Daniel: think it was done by Victor

McCool: let me reach out Victor then
... topic for the future meeting

Koster: have found some resource
... [Discovery and Software Adaptation using a Semantic API]
... (dated 31 August 2018)
... essentially, we have event, action and property as interaction affordance
... it's basically same semantic annotation
... let's go through this
... [Semantic API]
... enable applications to interact with connected things
... annotation by Thing Description
... like using English "go find a light ball!"
... you can make the filter successively finer

McCool: we should capture semantic filtering
... we need use cases
... let's go through this

Koster: realize the second narrow waist in system design
... complement to common network protocols
... sort of semantic narrow-waist
... for super portable applications
... this is the way of OneDM as well
... that's the whole idea
... including discovery
... and adaption
... [Semantic API Examples]
... semantic lookup returns instances capable of semantic lookup
... here is a template
... e.g., "iot:Light"
... resulting a TD
... and property definition here
... this is how APIs work
... want "iot:BinarySwitch" here
... turn on switch using that
... just semantic stuff here
... read the switch value by switch.read({"@type": "iot:BinarySwitchData"})
... may be adding ODF features
... mainly for classes
... the idea is oneDM is working on similar way to iotschema
... sort of like schema
... we only provide data element
... [Semantic API Examples (2)]
... RGB color example

McCool: basically, search by name
... also search by field using schema

Koster: yes

McCool: RGB is a good example
... could be an array
... or individual values

Koster: here is schema
... [Example Schema with Annotation]
... left side is "array"
... right side is "object"
... node.js sort of schema
... look at the schema and find payload by json pointer
... [Semantic API Examples (2) and Semantic API Examples (rivisited)]
... oneDM can access the data
... combine oneDM and iotschema

McCool: happen to know the annotation
... to see which one is using TD

Koster: right
... have to put it at the right place
... search for a right place to find the payload

McCool: may need to manually check the annotation

Koster: [DataInstance Dictionary]
... was working on that 2018
... [DataInstance Abstract Class]
... the idea is just providing decorator
... [Semantic API Examples (2)]
... this would be an array and you can run it on console
... a lot of this is already done by node-red, etc.
... node-wot is kind of underlying thing

McCool: more than one annotation?
... if you have different schemas at once?
... how we capture this idea?
... we have your presentation
... would like to have examples
... e.g., encode RGB light
... to make our ideas clearer

Koster: ok
... there are 4 things: discovery, adaptation to transfer layer, adaptation to schema and adaptation to data type

McCool: interesting

Koster: adaptation is interesting too

McCool: let's go back to your code example

Koster: [Semantic API Examples]

McCool: semantic search
... possibly multiple things
... might be OK to broadcast

Koster: think it's straightforward
... you might not know the original query

McCool: basically an item that matched
... that's allowed to access
... logistics here
... how to capture the idea and get into the detail?
... some summary document?
... let's go back to the agenda
... we need to finish the use case description for discovery
... let's brainstorm for that purpose
... (adds comment to the use case template)

Use case template

McCool: Filtering and Adaptation
... an example of light is good to express necessary complication

Zoltan: btw, this doesn't require a directory service

Koster: you have to have an index in that case
... could do that

McCool: one action is writing up a use case here
... ok to submit these descriptions on README.md?

Koster: make sense to include enums

McCool: (adds "enums")

[[

Semantic Discovery

Filtering and Adaptation

Use semantic categories such as those defined by iotschema or ODM to select interactions by type rather than name, as part of the discovery, e.g. filter parameters in a directory query. Within an interaction, select data fields by type rather than structure. Specific example: find an RGB light and extract data fields for each color channel independent of the name of the property name and structure of the payload (e.g. map vs array). Extension: additionally perform adaptation on units and scale, for instance lumens vs. percentages. Colored lights also raise issues of color space conversion, possible nonlinear mappings, clipping, lossy conversions, etc. A simpler motivating example would be a binary switch that adaptation still raises many questions on in the representation of the switch state (boolean, integer, enums, etc).

]]

Koster: we just put our model but should try more structured way

McCool: ok
... would assign the expected use cases
... (assign some of the use cases to himself)
... trying to get Michael Lagally for digital twins
... someone should volunteer to the remaining use cases, e.g., proxies

AOB

McCool: we should finalize the agenda for the virtual f2f
... will be held in 2 weeks

Virtual f2f wiki

McCool: (goes through the agenda)
... discovery topics
... 2-phase architecture (McCool) and DNS-based discovery (Chris)
... (adds "Use cases" and assign that to McCool, Koster, Lagally)
... and discuss previous work in WoT

Kaz: are you sure about "5min" parts like privacy and previous work?

McCool: previous work is just checking URL resources of the previous work
... on the other hand, we need to work on privacy work at some point
... (makes "privacy considerations" part of the "Two-Phase proposal" tentatively)

Koster: the question is where the TD comes from

McCool: we have to clarify that
... at the moment, we have several big categories for the agenda
... any other business?

(none)

[adjourned]