[MPEG-OTSPEC] Hints, TT and CFF (was: Re: Proposal to make OFF complete)

[Terence Dowling]

Typography is a complex subject with multiple domains of expertise.
Layout, shaping, glyph design and rendering all play a part. Asking
an individual to be an expert in all of these domains is not practical
or prudent.
Quality rendering is inherently device dependent. Consider the original
Laserwriter - 300 dpi with a spot size that reflected about 200 dpi.
Consider
the follow-on HP devices, still 300 dpi but with a spot size more nearly
400 dpi
and hardware processing to make smooth diagonals with the side effect of
turning an "O" into something that looked like an octagon. Consider the
Compaq laser printer 300 dpi in one direction and 600 dpi in the other
compounded
by supporting both long edge feed and short edge feed paper paths. Not
trivial to
get the result to be best possible in both orientations.Consider modern
color display
technology (some vertical stripe RGB and some more complex sub-pixel
components.
Color balanced color filtering isn't something a font program is well
suited to handle.

Asking a font artist to understand how to optimally move control points
for all of
the potential rendering delivery situations is a daunting task. Type 1
fonts and CFF
with competent rasterizers span this range. Getting Microsoft quality
results with
instruction based TT fonts requires Microsoft to take substantial
liberties with unspecified
instruction processing.

The Adobe model generally separated font production and rendering. Font
designers
were discouraged from altering hints to get specific results, rather
when there was an
issue the rendering code was improved to handle the evolving challenges.

Lets be clear, the font technology wars are over but just as Beta was
better technology than
VHS, but VHS won in the market so to TT has prevailed.

Terence Dowling
Adobe Font Rasterizer Development 1991-2010

On 8/26/2020 18:34, Eric Muller wrote:
> On 8/18/2020 9:58 PM, Behdad Esfahbod wrote:
>> The claim-to-superiority of CFF format is: 1. better hinting, and 2.
>> better compression. Re better-hinting, the interpretation of CFF
>> hints is NOT specified anywhere.
TT instruction processing as implemented in high quality rendering code
is also unspecified.
>
> In fairness, and without taking side on the proposal to remove CFF/CFF2:
>
> Although there is a spec for TT hints, it's hardly complete and
> requires quite a bit of reverse engineering if you want something that
> works. And there is more black magic if you want something that
> behaves like Microsoft's implementation(s).
>
> CFF leaves the interpretation of the hints to the rasterizer, with the
> intent that rasterizers can be improved without having to rework the
> fonts. So the specification is complete if font designers know what to
> do. I'd be interested to hear from font designers: do you think you
> understand how to set hints in CFF fonts? Are there things missing
> from the Type 1 Book + Type 2 technical note?
>
> Eric.
>
> _______________________________________________
> mpeg-otspec mailing list
> mpeg-otspec at lists.aau.at
> https://lists.aau.at/mailman/listinfo/mpeg-otspec