The change of semantics of score.txt is a bit subtle

[niemela]

Comment from @gkreitz:

The complete change of semantics for score.txt is gonna be a complete pain to deal with (it's either a float in [0,1] if the test group score is bounded, or the actual score). It would have been much better IMHO to separate those two modes of operations into two files (score.txt vs scaledscore.txt ?) to not break every legacy problem

Thoughts on this?

[Tagl]

I think separating is fine. Might make auto-updating packages easier too.

[eldering]

I have no strong opinions on the matter, but I have to say that when I was reading the specification it was not very clear due to the complex inter-dependence of "unbounded" and scores in subgroups, etc. So I think that indeed separating this will also make it clearer.