Meeting minutes
Guests
<EgeKorkan> https://
Patent Policy because we have guests from the Autonomous Agents on the Web CG
Ege: (describes the notes around the W3C Patent Policy)
Minutes
Ege: can we review the minutes tomorrow?
Kaz: that is good
Ege: no objections, will be reviewed tomorrow
Manageable Affordances
Ege: here are they from webagents cg
https://
Cristiano: I will also give an overview
… I want to have an open discussion. please share feedback
… this work item started some time ago
… like actions that can be queried after starting
… I am now looking into iot platforms
… surprisingly, it is not that common
<kaz> Yet another Action model
Cristiano: I have collected a first set and I am going through them in more detail
… we also have proposals from some time ago, including one from me
… there were proposals to change TDs when action starts, use new terms in action level, and mine was about using TMs for the action modeling
… do you know some platforms that have this feature
Danai: what are the issues with updating TDs
Cristiano: can someone explain the reasoning?
Ege: you can go ahead first
<kaz> analysis-manageable-affordances.md
Cristiano: avoiding overfetching TD, having static TDs to program against
Daniel: there were security concerns
Daniel: also if the TD keeps changing, it is difficult ensure proper signing
Ege: td changing means updating the td in TDD and in the Consumer. Also the way to program the application program changes, requiring adaption on consumer programmers or the implementations
… also it is overkill if only an action's availability of operations change
… might make sense if the whole action dissappears due to state change
Cristiano: do you see this pattern in agent systems
Danai: we work with the assumption that the TD can change. Agents are expected to continuously observe the TD
… so this removes the obstacle of updating the TD
… the alternative I have seen is that when the agent starts the action, a new resource is created that can be used again
… maybe others can comment
Cristiano: creating the resource goes a bit in the direction that I mentioned with TMs
Cristiano: How does the agent get the action resource to be used later on? is it a protocol in the end?
Andrei: for us it is about hypermedia APIs. We consider them dynamic
… the semantic model is hardcoded in the action
… if a new model of an action appears, the agent ignores it
Cristiano: we also need to consider the case of seeing actions created by others
Andrei: there are also research on putting pre and post conditions for actions
Ege: there are overlaps with TM-based approach and the work from Danai and Andrei
Ege: we should look into the physical case, so what is happening in real world
Kaz: it would be interesting to explore state machines
… it would be interesting to combine that with TDs
Cristiano: how can they express such state machines in a WoT-friendly way
Ege: there are always some state machines, since Volkswagen, etc., are already using state machine as the main controller to organize multiple agents for smart car purposes.
Cristiano: would there be a common denominator?
Koster: how to design a hypermedia state machine standard
… that is our question
… how to build it in an explicit way
Cristiano: If I dig into the bacnet alarms and the microscope example, I could find more uses
<cris8> w3c/
Ege: I think we all agree on trying to describe existing mechanisms
Andrei: we are starting to wrap our it
Cristiano: this hypermedia state machine can be something to work on together
… seems to align with the TM idea I had
Ege: I think we can meet again towards the end of august or beginning of september
(adjourned)