Per our previous discussion, I think the issue is really that it is a spec of specs and neither part of that is followed. I don't think it's actually useful to discuss dvms as a spec but it is useful to discuss the use cases nested within. The same way a company isn't going to build all their doors bigger just for the vending machine, a client shouldn't re-engineer the client to fit a specific type of DVM in. Does the transcribe DVM fit into your client? Then implement that. If the spec isn't followed, pick the one you prefer and use it and prompt that person to PR that part of the DVM repo. Do that for 2 more dvm types and then a pattern will probably emerge that will make you realize what the actually problem is with the spec itself. I don't think there will ever be a client that can implement all dvm types, but there will likely be pockets that end up specd out in ways that can be.