Re: [ft-devel] State of TrueType interpreter v38 and up?

[Werner LEMBERG]

> I thought one of the core ideas of ClearType is that glyphs are
> snapped to the pixel grid only on the Y-axis?

No.  This is what *effectively* happens, due to the strongly increased
horizontal resolution.  However, at least in `native' mode, you can
still enforce non-integer positions along the x axis.

> By Y-only hinting I mean what the slight autohinter and the CFF
> engine do.

Such a mode doesn't exist in the MS engine.  However, Adobe's modified
TT engine essentially hints only along the y axis, if I've understood
Dave Arnold correctly, replacing x axis hints with emboldening.

> So what exactly must the TT driver support or imitate so that it
> fits the description of “Use native vertical-grid-only-snapping if
> driver and font supports it [...]”?  If I understand correctly, FT
> already does subpixel positioning and natural advance widths in
> light mode (autohinter, CFF)?

It must

  (a) ignore changes of the horizontal advance widths due to
      hinting instructions (not implemented yet), and

  (b) ensure that all instructions behave correctly even if the
      horizontal resolution is increased (this is ClearType
      compatibility mode, already implemented).

>> The latter: MS uses super-sampling, which *virtually* increases the
>> device resolution by a factor of 10 or so, while FreeType does
>> *really* increase the resolution.
> 
> I take it those differences don't matter?

Well,...

> Any known advantages and disadvantages to either approach?

... the MS approach is superior IMHO.  It

  (a) elegantly circumvents many problems due to the extreme
      difference between horizontal and vertical resolution – it
      basically makes the compatibillity mode redundant, and

  (b) the increased resolution is around 10 times in both horizontal
      and vertical direction, compared to a threefold horizontal
      resolution increase of FreeType.


   Werner