Host automation based controllers ultimately have a higher ceiling of potential than MIDI. There's a narrow category of DAW-specific hardware - thinking of Maschine, Push, Reason stuff - that shows this. However, for something that is not tied to a DAW and in fact implicitly supported 3rd party developers, VST2.x is pretty much the primary/generic consideration.PAK wrote: Even with their flaws, host automation based controllers are the way to go for anyone controlling computer plugins!
VST2.x has pretty limited host automation and really doesn't consider hardware control at all. In the context of where things were at 15 years ago when it appeared, that's understandable, and it's obviously been a successful standard nonetheless. Meanwhile though, the assumptions have changed. Plug-ins, DAWs, and hardware controllers have grown considerably in variety, scale, and complexity, past what VST2.x was probably intended to handle. (Tellingly, there's a class of bugs and work-arounds related just to the number and order of VST parameters - I suspect DIVA's per-voice detune controls might trace back to this).
One might dream ... if the default plug-in specification was more conscious of hardware controllers, and did that well, everyone might benefit. Users might expect what's really kludgy in Automap - building control mappings, getting the pages and the ranges/resolutions right - to be really quick, perhaps automatic or creatively intelligent in some cases. Any hardware manufacturer could support it pretty directly, in any DAW. DAW and plug-in developers could offer the feature of really smart hardware integration for a very reasonable investment of planning/coding effort. (Some host automation issues that aren't directly related to hardware control would also be resolved, I think.)
Sadly, 2013 suggests completely positive advances in plug-in standards might be a ways off. VST3 has crippling conceptual flaws making host/plug parameter stuff a bit of a clusterfuck, Apple's new security model is a clusterfuck for AUs. AAX? iLokalyptical clusterfuck. I'm worried there's at least another half-dozen or a dozen clusterfucks in the way of a plug-in standard that's a decisive improvement here. Despite the long-term potential, in the meantime I think MIDI is worth thinking about; not saying it's completely cluster-chaste but it's possible and more of the pieces are in place that it might seem
(I'm also reminded of a maxim - all technical approaches to a problem eventually converge; maybe part of what a better MIDI might look like is also part of what a better VST standard would look like)
[e] I really think adopting a little bit of logical typing and metadata for parameters in VST-like formats would go a long way here. There's a danger of trying to do too much, or missing the mark conceptually, making the back-end engineering infeasible. But maybe it's analogous to how the first versions of HTML really didn't add too much to a plain-text version of the same content, but still were completely revolutionary. It's still difficult to see any time soon, IMHO.