The McLaren Synth Kit presents audio synthesis as the construction of a graph of audio operators (voices) with the parameters of each type of operator determined by an associated model. Voices define the code that is run in the audio thread and models hold the values that are read from the audio thread. This separation allows the audio thread code to be ensured that maniplations on the models do not block. With models safely out of the audio loop, UI interactions with model properties is straightforward and can use Cocoa idioms for manipulating property values.

The McLaren Synth Kit (at https://github.com/mclarenlabs/libs-mclaren-alpha) is an Objective-C toolkit for building projects that simplify working with MIDI and sound synthesis on Linux. We open-sourced it last year and have been refining it with new features and capabilities.

A few years ago, I explored the idea of writing a small interpreter for embedding into McLaren Labs applications. The concept was to extend rtpmidi or a synth application with a scriptable ability to customize controls, behaviors or sound graphs. This would help make each application very slim, but to allow extension through scripting. (This is not a very novel idea: it's been done many times before. :-) )

There's a cool project called "GNUstep Desktop" that brings together many old GNUstep technologies, and a few new ones, to provide an entire integrated desktop. We gave it a whirl and are pretty impressed!