Does VSS-syntax only support a single root?

[erikbosch]

Today we mention here and there in the documentation that the VSS catalog forms a tree, and vss-tools actually gives an error if you have more than one root. But we do not really say anywhere that a model using VSS-syntax must have a single root, i.e. that something like below is an invalid VSS model.

If we want to keep the idea that there can only be a single tree and a single root I think we should improve documentation so it clearly states that. If we do not think that is a reasonable limitation we should start looking at how vss-tools can support multiple trees.

(Whether a VSS model must have exactly one root or not does not necessarily affect downstream implementations, as an implementation may support access to multiple models in parallel)

a:
  type: branch
  description: Branch A.

a.b:
  datatype: float
  type: sensor
  unit: km
  description: A sensor.
  
c:
  type: branch
  description: Branch A.

c.d:
  datatype: float
  type: sensor
  unit: km
  description: A sensor.

[ppb2020]

FWIW, in our implementation, we do rely on the fact that there is a single root. It probably would not be a big deal to adapt,but that would require a few changes, some of which may not be trivial.

[UlfBj]

If a structured and comprehensive model that supports multiple roots/trees is desired, then there is HIM.
https://covesa.github.io/hierarchical_information_model/
The VSS rule set is a subset of the HIM rule set, so switching to HIM is in principle done by pointing to the above documentation instead of the current.

[erikbosch]

Similar to the discussion on node/signal names (mandatory rules vs. recommendations) we could actually think of a dual approach:

• VSS tools (and VSS syntax definition) shall support multiple roots
• But it is recommended to only have 1 root, and that would be a requirement for the standard cataalog (i.e. with --strict you would get an error if having multiple roots)