Annex A "Diagrams", whilst /informative/, sets out a confusing definition of the diagram types.
On the one hand, there are nine SysML diagram types that have "model elements and corresponding concrete syntax", with each diagram kind cross referenced to the specific clause defining a specific part of the SysML modelling language. So for example an activity diagram "represents" (by virtue of the appropriate nodes and paths) the model elements and connectors described in the Activities clause (clause 11).
On the other hand, there is the weaselly statement "Although the [diagram]taxonomy provides a logical organization for the various major kinds of diagrams, it does not preclude the careful mixing of different kinds of diagram types". A similar remark can be found in UML 2.5 in the form of a note in it's Annex A "Diagrams" section:
"NOTE. This taxonomy provides a logical organization for the various major kinds of diagrams. However, it does not preclude mixing different kinds of diagram types, as one might do when one combines structural and behavioural elements (e.g., showing a state machine nested inside an internal structure). Consequently, the boundaries between the various kinds of diagram types are not strictly enforced."
So, despite Annex A being informative, it creates at least two possible interpretations for tools to implement:
1. A diagram can present representations of any model element and connector, or
2. Only the specific model elements associated directly with a diagram kind (according to Annex A) can be represented on a diagram with a specific kind.
Case 1 would allow all manner of diagrammatic mixing of model elements from the different semantic clauses, even when these clauses do not allow for any logical connection. However, it might allow an activity diagram to show a «Block» with an owned activity model. It would also allow a use case diagram to include a «Block» connected with a communications path to a use case node (since neither the use case or block clauses seem to set out a constraint to prevent this in the model).
Case 2 would only allow the model elements from the directly associated clause along with the model elements that are common to all diagrams. This is a strict implementation, but may have consequences where more flexibility would have been desired. For example a modeller using Blocks, or user stereotyped classes, for stakeholder modelling would find that they cannot use these model elements directly on a use case diagram where only use case and actor model element types can be represented.
The standard should also consider the practicalities of modelling tool design and usability considerations. Case 1 would imply just one diagram implementation and a large model element library (or tool bar) for the user to work with (in a model creation through the diagram paradigm). Whereas, case 2 requires distinct implementation of the diagram kinds along with their implied constraints of which model elements can be added to them, although it does mean that the model element library (or tool bar) is quite specific to each diagram kind.
Without standardising what each diagram kind can contain, there will be no possibility of model interchange that includes the diagrams.
On a final note, the explicit exclusion of the Profile diagram from UML seems strange because defining and working with Stereotypes is a common (albeit advanced) modelling concept in MBSE, but to do it requires specific modelling concepts (and their associated nodes and paths). Hence, the current suggestion that this is done with package diagram would seem to suggest that the aforementioned diagram case 1 was being implied.
Suggested change (with a strict diagram syntax leaning):
1. Make Annex A normative and remove the diagram mixing statement: "it does not preclude mixing different kinds of diagram types"
2. Review each of the modelling clauses 8 through to 16 to check that they completely define the set of model elements and connectors that are useful to make available to modellers on the associated diagram kind. That is make explicit what is otherwise inferred by the lack of a constraint. E.g. that use case model elements can (or cannot) be connected to Blocks (or other forms of stereotyped classes) with a communication path connector.
3. Add the UML Profile diagram to the list of SysML diagrams.