Authors specify their window size in CEA708 in number of characters (assuming a monospace font - otherwise it's the number of widest characters of the font).
Specifying the region width by percentage doesn't allow the region to grow with the font size.
This has to be fixed.
At FOMS we discussed to use number of (widest) character as the width count instead.
Since regions are not implemented anywhere, we can simply make this change and not break backwards compatibility.
Maybe we should allow floating point percentages and ems everywhere, for simplicity?
I think region width might be best in number of widest characters. But we could add ems also to cue width if that is desirable.
It's not hard to come up with a hypothetical scenario where a cue (without a region) should be just big enough to hold its content, in which case a font-relative size is required. But my real argument here is of course consistency/simplicity.
I agree with Philip that it might be best to add it to cues as well for consistency. That way authors have a choice between using ems for the entire VTT file or percentage for the entire VTT file and not have to flip back and forth.
Fine by me.
*** Bug 23849 has been marked as a duplicate of this bug. ***