Comment 6 for bug 1617466

I investigated this little bit. In general form XSPF/JSPF doesn't do what we want. Although it has extension system what VLC uses to extended XPSF (https://wiki.videolan.org/XSPF/) which gives at least second start and stop but again it's not as accurate as we need.
MIDI could be the one but problem it is binary and mainly it's formed more or less '80. There is XML Midi (http://www.palserv.com/XMidi/ ) and musicXML (http://www.musicxml.com/for-developers/).
XSPF extended with musicXML could be one option because we want to map what has happened between starting to play song and stop when it have been mixed to another and again record what happened to that new song.
Binary MIDI would be best option but if one wants to read it or see whats inside it's pain.

One option is adopt: XML that are compatible with: MLT framework (https://www.mltframework.org/bin/view/MLT/MltXml) so one can tune mixes and then render it offline.

For people who just requires played songs we should use CSV file. Easiest to parse and use with Word or Libreoffice. Just couple ideas to get this done.