Comment #10 on issue 2707 by [email protected]: Patch: Midi volume fixes http://code.google.com/p/lilypond/issues/detail?id=2707
The semantics have not changed in any manner that produces indeterminate output
For at least two years (maybe forever) Lilypond has set dynamics independently for each Voice, and if the user does not start a Voice with an absolute dynamic (\p or \f, etc.) Lilypond arbitrarily starts the Voice at a mid-range dynamic level. With Heikki's patch, the 'mid-range' level takes equalization settings into account, and you get a warning if Lilypond makes this arbitrary setting of midi dynamics.
