Dependency arrows from provided to required I/F

[mse]

Why does Enterprise Architect allow drawing an arrow from a PROVIDED interface to a REQUIRED interface? I thought a dependency arrow can only go from a REQUIRED interface to a PROVIDED interface because a required interface depends on the provided interface. However, EA Architect supports drawing the dependency arrow in either direction. This seems like a violation of the UML meta-model. Can someone clarify why this is the case?

[qwerty]

I think you can also tell that the definition of the provided interface depends on the required. Means: you need to implement some interface so I can use it. The dependency does not tell anything about usage/data flow. If you use the lollipop connector (can't recall the name) it's always drawn the way you would expect.

q.

[Geert Bellekens]

In UML a dependency is defined as a relation between two NameElements.
I haven't seen any constraints that indicate you couldn't model one from a  provided to a required interface.

It might not make sense, but I don't thinks it's invalid in terms of UML syntax.

Geert

[mse]

True, I never read it from the perspective of a provided interface depending on the existence of a required interface. I agree, it doesn't quite make sense and I have rarely seen it drawn with that intent. Rarely would anyone say the following:

I have a component "A" that provides services. Component A however needs (requires) component B to subscribe to component A's services. Consequently it has a dependency.

Perhaps that is the advantage of using the assembly notation.

[qwerty]

I think it makes sense. A company does not (better: should not) provide an interface just for fun but because the customer requires it. So the provided depends on the required interface. Of course we all know companies that offer interfaces because they think they can persuade their customers to use them: Use This Interface!

q.