Feature rule set extensions

[alanmapleburl-au]

These extensions to the VSS allow consumers to understand each signal well enough to determine its relation to signals emitted by vehicle components and to downstream client and consumer needs. Because they are optional extensions permitted by VSS Overlays, they do not break any current code or processes.

The properties.md file describes the promotion of dataProperty and objectProperty to a first-class types, allowing signals to reference the property they report.

The elementType key provides each entry's classification within the vehicle signal domain, such as SignalDefinition, FeatureOfInterest, Property, etc.

The identifier key provides a unique and immutable key. This allows restructuring of the tree and the files/folders without any effect on the key. In my examples I've used the entire path of the entry, but it could be any unique value.

The definition key provides the necessary and sufficient conditions of the entry which can translate to formal axioms.

The featureOfInterest and property keys in the attributes, sensors, and actuators explicitly refer to the object being observed or manipulated, and the property being reported. This is needed not only to aid integration, but to allow analysis, discovery, automation, and inferencing through the graph.

[SebastianSchildt]

I made some small comments, but generally, to get a better understanding, can you somewhere upload how the standard catalogue would look like implementing these suggestions? I.e. modified vspecs or something you generated. I guess like all things, what specifically to use as content for your suggested properties might be discussed on a case by case basis, but seeing some "this is how I meant it" example of the VSS as you have modified might help the discussion.

[erikbosch]

As general info to reviewers - this PR is related to the presentation "Integrating Vehicle Signals with VSS and Metadata" which is available in the COVESA wiki. Please review the presentation to get more background information.

Some areas that I think we should discuss:

• There have been similar discussions in the VSSo group, are we ready to "promote" the proposed extension as official (but optional) extensions, or should we better for now look a how this can be managed as a custom VSS profile, and how VSS and vss-tools can manage custom profiles.
• If accepted as an "offical extension", how do we want the VSS standard catalog (the *.vspec files) to use the new additions. Not at all, or do we gradually want to introduce some of them, ans possibly long term replace some of them with the new concept? Maybe we need a table or similar to state what we expect or accept in the standard catalog for new signals.
• How can we define semantic rules, like if a signal has specified datatype "uint8" and property "MyProperty" which in turns specify "uint16", what does that mean? Is that a semantic error or does someone of them have precedence? Or is it up to the user?
• How do we want vss-tools to handle the new items. An obvious question is how/where properties should be defined. But apart from that - how shall output for existing tools be affected? Vss-tools may for example generate identifier, and it is easy to include in yaml/json output, but not obvious how to include it in other formats.

[erikbosch]

Meeting notes:

• Erik: Adnan&Erik to have meeting with Alan to discuss way forward
• Peter: This makes VSS difficult to understand, would be difficult to convince people to use it.
• Daniel: Important to specify what requirements this solution fulfills. We need a framework to define and discuss requirements on VSS.
• Erik: But we do not have any formal requirements on VSS as of today, could possibly be useful if managed well
• Daniel: There have been some discussions within VSSo, but Ford presentation cover many ideas not yet socialized in VSSo. I see values in some features, not all needed
• Daniel: we need COVESA guideline on how to document projects/initiatives, a simple form for each project. We need to make PRs more objective, the list of requirements must be shared/agreed. I have offered Paul to assist in establish guidelines.

[alanmapleburl-au]

@SebastianSchildt Please see Erik's general info to reviewers. He refers to a presentation that will satisfy your desire to see it in action. @erikbosch I would follow up the acceptance of this PR with a series of PR's that populate the additions @erikbosch I think the initial rule on signals is that the property's data type, unit, min, and max are reviewed for any overrides. That doesn't break anything. Eventually the rule could be that those k/v pairs default to the specified property if omitted. @jdacoello I tried to specify the requirements. If you comment the ones that seem unnecessary I can elaborate.

I'm interested in the notion of a custom VSS profile but couldn't find any info on that.

[jdacoello]

• Daniel: ...I have offered Paul to assist in establish guidelines.

Here is the first draft for such a methodology. I explain there the fundamental components that must be described by each of the COVESA projects and its associated artefacts using a unified structure.

https://wiki.covesa.global/pages/viewpage.action?pageId=85196928

[erikbosch]

I spent some time thinking how we possibly could "transform" this PR into a "VSS Profile". Not straightforward as we do not really have anything called "VSS profile" as of today, but I created a draft PR to show my ideas. My idea is to create a section in the documentation where we document more or less official VSS extensions. In my draft I created something very short for the Eclipse KUKSA DBC mechanism and what I consider mostly as a placeholder for the "Ford Data Properties" (or whatever we would like to call it). This could possibly be a way forward to capture the good ideas from Alan but at the same time postponing the decision on whether this shall be an official/recommended profile or even integrated into the official syntax.

[SebastianSchildt]

I like the general direction, as a word could we maybe rather call this "VSS Extensions" or "VSS Addons"? maybe somebody has a better term, but for me "Profile" either sounds like "Two wheeler profile" "Truck profile" pr potetnially something like "VSS base profile", i.e. in my head I connect it more with the "catalogue" part, whereas the things suggested here are bit more about adding new "features"/functionality to the model

[erikbosch]

Meeting notes:

• Daniel: Risk if we allow addition of extensions that we split efforts, that it will not kept in synched with other efforts

• Daniel: Would recommend a clear scope of what the VSS Yaml supports.

• 1. Level of agreements - terminology/vocabulary, schemas
• 2. Tooling

• What can be done - right now targeting vocabulary

• If we want schema we may target a more expressive/modern language to express it.

• Erik: Is conclusion that proposal shall be seen as input to VSSo project, and not included into VSS documentation as of today, neither as std or as "extension".

• Daniel: We should extend scope of ontology. VSS project should focus on vocabulary of VSS signals. We should better use GraphQl tooling so that we do not need spend so much time on tools. Better than expanding VSS syntax.