Hazard Annotation

[lu-zero]

Use case for SIFIS-Home Hazards in WoT

[mmccool]

Would suggest making the title more general, e.g. "Home Hazard Annotation", and then just link to SIFIS internally (e.g. describe the use case, not the specific application or product).

[mmccool]

Also, could be generalized to "Hazard Annotation", e.g. it could be extended to cover hazards in an industrial or smart city setting as well. That may or not make it too broad (to discuss). My personal feeling is that it could be expanded to cover these situations without significantly expanding the complexity of the use case itself. A privacy risk is a privacy risk whether it is in the home or in the city.

[lu-zero]

Thank you for the review, Friday I'll fold all the change and rebase it as non-draft :)

[mmccool]

Comments:

• Should probably better define the problem, and focus on that rather than assuming a solution.
• We may also want to define sub-problems: e.g. defining hazardous states, identifying interactions that can cause hazardous states, sending notifications when a hazardous state is entered, defining policies preventing hazardous states being entered, identifying contexts where the hazards exist
• Defining some of these may require coordinating multiple Things. For example, you might not want to start and warm up the car if the garage door is closed, in the context where the car is in the garage. We should probably state that as a requirement (we can figure out how to solve it later).
• Title is still not quite right, since it presumes the solution. Perhaps the problem is "Hazard Identification and Prevention"

[mmccool]

If we want to define Hazard categories, e.g. for annotations (fire hazard, chemical hazard, etc) should reference existing standards, for example, https://www.ccohs.ca/oshanswers/chemicals/whmis_ghs/hazard_classes.html, ISO standards, etc.

[lu-zero]

Comments:

* Should probably better define the problem, and focus on that rather than assuming a solution.

The problem I'd focus on this specific use-case is adding the information, I can expand to the full picture, but it feels a bit too wide.

* We may also want to define sub-problems: e.g. defining hazardous states, identifying interactions that can cause hazardous states, sending notifications when a hazardous state is entered, defining policies preventing hazardous states being entered, identifying contexts where the hazards exist

I can put it in this use-case but I'd consider it dependent from this use case and separate. Policy definition and enforcement, at least in SIFIS is a different topic. Same for ongoing danger identification and reaction:

• Hazard labeling provides the "ground truth". Its main purpose is to make the actors aware of the risks and that's it.
• Policy definition and enforcement are to prevent the hazards from happening and in general control the behavior, potentially emergent.
• Detection and reaction are in place in case the policies are lacking or the enforcement somehow failed.

* Defining some of these may require coordinating multiple Things.  For example, you might not want to start and warm up the car if the garage door is closed, in the context where the car is in the garage.  We should probably state that as a requirement (we can figure out how to solve it later).

That is a policy, potentially implicit/embedded in a Garage Thing that consumes its door and a CO₂ sensor. In the SIFIS-Home scenario we have a separate actor that collects all the polices and is queried before any change to the state of the whole Home happens.

* Title is still not quite right, since it presumes the solution.  Perhaps the _problem_ is "Hazard Identification and Prevention"

I can go this direction even if that means making the use-case much wider.

If we want to define Hazard categories, e.g. for annotations (fire hazard, chemical hazard, etc) should reference existing standards, for example, https://www.ccohs.ca/oshanswers/chemicals/whmis_ghs/hazard_classes.html, ISO standards, etc.

Chemical and biological hazards have good categorizations, the other kind of hazards that are closer to connected devices seemed to not have the same level of detail. For the purpose of SIFIS-Home we developed a tiny ontology to showcase the feasibility, but if the use-case is deemed worthy we could to contact probably the people at SAREF or such organization to see what are their plans in this regard.

[lu-zero]

I tried to expand adding few more scenarios and I covered a bit the case of aggregating information to produce an evaluation of risk.

[mmccool]

Discussion from meeting:

• Scope was expanded (e.g. policy and not just annotation)
• Was made less specific as to the solution
• Open issues: need ontologies for hazard categories (need broad categories; SAREF?)
• Regulatory approvals?