Well, seems that this kind of problem is very common here (check this, this other and mine)
and not only (here) and here @Ardour), but at the moment this request is almost unattended.
BTW, after some intense discussions, NASPRO project is born: it will be "a free, modular real-time capable, thread-safe, scalable, standard-agnostic, cross-platform DSP framework aimed at becoming a powerful and reusable sound processing engine.
It will be made of a core library, named NASPRO core, containing features needed by the rest of the framework such as thread-synchonization mechanisms, an OS abstraction layer, error handling, message reporting, etc; then there will be a set of different libraries, modules, and maybe applications, each facing a single aspect of sound processing.
At the moment the only library planned, obviously after the core one, is the NASPRO objects library. This will handle the purely functional part of processing objects in a very generalistic and modular fashion, so that these can not only be regualar processing plugins in different formats (such as LADSPA, DSSI, LV2, VST, AudioUnits, etc.) but also everything else (un)useful for sound processing (mathematical formulas, circuits, neural networks, random data, etc.)
At least one collection of modules for the NASPRO objects library will be released as well, containing all the code needed to make different classes of already-existent plugins work."
Official website
Well, actually seems the best approach to have (when will be released) a multiplatform open source multitrack recording/editing software.