Probably I'm missing your point, but the fundamental dynamics equation of a manipulator describes physical relations among quantities irrespective of the particular electronics/mechanics running underneath.
If there is a rotating shaft with a joint mounted on it then there is always an acceleration. In a stepper, velocity and acceleration might be significantly discontinuous due to the nature of the actuation, but they can be still computed/measured/estimated.
This harsh behavior of such fundamental dynamic quantities is one of the reasons why DC motors (+ circuitry) are preferable over steppers, with the cost of higher expense.
Moreover, to complete the reply, a stepper is usually employed in a open-loop configuration, which motivates its cost effective choice. Therefore, if the datasheet does not provide the user with a model of how the stepper will accelerate/decelerate under some given input conditions, then we are somehow forced to measure with a dedicated equipment what is happening at the joint side (e.g. see this paper) to be able to retrieve information about the ongoing dynamics. Clearly, this goes against the initial idea of cost saving and in this respect other solutions are more effective, as previously said.
Further, it is well known that feed-forward models themselves are inaccurate description of the reality, thus feedback is always required, especially in dynamics context.
To conclude, my opinion is that steppers in a open-loop scheme are inadequate for dynamics control.
By the way, here's recapped for your convenience the model of the stepper: link. You have a matlab block for that you could employ to experiment; the model is the same cited in the paper above. In principle, if you find estimates for all those electrical and mechanical quantities, you can consider going open-loop.