🚀 go-pugleaf

In article <1992Nov20.020424.5847@adobe.com> mmwang@adobe.com (Michael Wang)  
writes:
> As the point size gets smaller than this optimum size, reading
> comprehension begins to suffer because it becomes more difficult to
> discern individual words and letters. Conversely, as the point size
> gets larger, reading speed suffers since instead of reading groups of
> words at a time, you start to focus on individual words or letters.

I get what you're saying and it makes a lot of sense.  I'm
just wondering how the optimum size is related to size
(or, I guess it would be better to say "extent") and
resolution.

If the extent of a letter is too small, it's unreadable no
matter what the resolution is because you hit the wall.  If
the number of pixels in a character is too small, you can't
read it either.

Let's not worry about resolution for the moment.  We can
avoid that by assuming we're talking about billboards or
very high resolution printer output.  The extent of a
character on a billboard might be the same as that on a page
at the expected viewing distance for both.  Imagine it's
not foggy out or anything like that. :-)

OK, now I think know about what extent is difficult for me
to read because it's too small.  I also think I see where it
gets too big.  

> To compensate for these problems, type designers in the past created
> slightly different designs for different point sizes. At smaller
> sizes, generally the x-height was enlarged, contrast between strokes
> decreased, width of the various letters increased slightly and the
> letterspacing was looser. At larger sizes, the opposite was done. The
> effect of all these changes was to try and make those sizes more
> readable.

Those all work for me, but it really looks like you're just
avoiding the bad sizes.  Actually, that's probably the
best solution. :-)

In Type 1 fonts, all the hinting is aimed at resolution
limits and it seems to work OK.  What type of hinting
mechanism would help optical scaling?  It'd be based on
actual size (or extent if we know it), right?

I suppose if you had a function that gave xheight as a
(nonlinear) function of pointsize and another for
letterspacing (sidebearing?), that'd be a good start.

Instead of a function, maybe you could just describe the
nonlinear regions somehow.  There ought to be a big linear
region in the middle and then a tapering off at both ends,
right? 

---
save /- 1 string def[331 64 779242 46 396 46 871217102 46 14782 10]{dup
99 lt{- 0 3 -1 roll put -}{32 =string cvrs}ifelse print} forall restore