Do you want to automate it? Perhaps with a noise profile or just select it manually and use a silence command. I like the cool edit pro software now called adobe audition.
You can also look at the spectograph and draw an delete unwanted frequencies within a timeframe.
Perhaps Imagination Studio would be a good investment as an alternative software package?
For osalt preferrers Ubuntu Studio comes with decent software like ardour audacity and vokoscreen and now cheese. It seem to be better suited for webcasting.
I like the functional pseudocode algorithm idea you have here:
I want to get rid of (completely silence) any sound shorter than 100ms exceeding a general silence of -48dB
I want to learn that too. I think you need to know how the audio is encoded before you can start dividing. Segments into 101ms does your software auto select the beat with arrow keys to start of next note? Perhaps Autokey or Autohotkey can help but you would need a way to check the sound length. Or manually sit there and press next like f3 finds