Comment 13 for bug 727733

Malcolm: yeah, (3.) which is 560×1120 is the one I've been wanting to knock up here at the weekend and test for practicalness to see if it's possible. If the terminals on GNOME/KDE/Mac OS/Win32 will all accept and use an WinDescent/WinAscent totalling 1120 with the em-size remaining at 1000 then this would give a better factual basis for making a decision.

The same work to bring the outriggers into that 560×1120 vertical bound /will/ still need doing, but it would then have the same x-height and cap-height as the proportional.

The question about the metrics is because there's currently inconsistency anyway, and I have no idea which set is getting used as a result. Here's the three groups for the original Proportional regular and the original Mono regular (nominally the same vertical size, but a spread between 1017 and 1127):

  >>> 932+189, 932+189+28, 776+185+56 (Proportional: Win, Typo, HHead)
  (1121, 1149, 1017)
  >>> 932+195, 776+185+56, 776+185+56 (Mono R21: Win, Typo, HHead)
  (1127, 1017, 1017)

If option (3.) works for the monospace (1120, 1120, 1120) then the proportional might for consistency want to be (1160, 1120, 1120)... the 1160 being the design extends of the double accents in the regular, which would be knowingly brought into 2:1 for the Monospace.