[MPEG-OTSPEC] Some research on TT instructions and cubics

[Skef Iterum]

Sorry for our confusion over the IUP instruction, folks. I guess one of the
risks of working from documentation alone is that you'll over-read it.

It appears that at the instruction level there are just points, with no 
semantic
information about kinds (on- or off-curve, etc.). Seems like that means 
that it would
be not that hard to fake up a manual hinting system -- maybe start with an
existing manual hinting setup, feeding the data through an interpreter 
to modify
the point locations, and then feed that through a rasterizer modified to 
deal with
cubics. (I assume Behdad or someone at Google has some form of the 
latter, or
I guess you could just convert to SVG or skia's internal representation.)

Setting something like this up and trying some manual hinting of a few 
glyphs
might be the quickest route to confidence about how cubics will interact 
with
TT hinting. (Autohinting will presumably take a while longer to sort out.)

Vaguely related: Following along the potential alternative of having the 
variable
component data in a separate table, we have started discussing what all 
this would
mean for CFF (e.g. PostScript lineage) style hinting. It's way too early 
for any real
documentation on this, but we have notes on what looks like a pretty 
good model,
which will require some changes to the implementation but can use the 
existing
font- and glyph-level parameters. If anyone wants early info on this we 
can talk.

Skef