Path: utzoo!utgpu!water!watmath!clyde!att!osu-cis!tut.cis.ohio-state.edu!mailrus!purdue!decwrl!ucbvax!betelgeuse!carlson
From: carlson@betelgeuse.uucp (Richard L. Carlson)
Newsgroups: comp.text
Subject: Re: Various article threads.
Message-ID: <26524@ucbvax.BERKELEY.EDU>
Date: 22 Oct 88 03:49:27 GMT
References: <704@umb.umb.edu>
Sender: usenet@ucbvax.BERKELEY.EDU
Reply-To: carlson@betelgeuse.UUCP (Richard L. Carlson)
Organization: University of California, Berkeley
Lines: 45

In article <704@umb.umb.edu> candy@umb.umb.edu (Karl Berry.) writes:
>
>Some comments on various articles:
>
>3) Although it is reasonable enough to worry about disk space
>from a computer scientist's vantage point, it is not reasonable
>from a type designers. The fact that Metafont outputs a different
>file for each type size is good, not bad. That means that
>fonts are not linearly scaled.

I agree that allowing different font sizes to have different
characteristics is a good idea.  For example, compare "cmr5" with
PostScript's Times-Roman at 5 points; I think "cmr5" is much more
readable.  However, I'm starting to think that the Metafont-style
of generating fonts is a little bit too restrictive.  A certain
finite number of font sizes are compiled and available for use;
there is no way to use other (often meaning larger) versions of
these fonts.  This problem is particularly acute for the several
fonts in the "cm" group that are defined only at a 10-point size.

Sure, you can scale the fonts up by some magstep to get a larger
font, but  (1) there is usually a rather small maximum magstep
value (magstep2 on our local machine); and  (2) we're back to
linear scaling for these larger fonts.

I think the optimal situation would be to take advantage of the
best features of both systems:  allow fonts to be arbitrarily
scaled, but allow fonts (especially at small point sizes) to be
scaled non-linearly.  While we're at it, why not allow certain
features of a font (such as some serifs) to be declared "optional"
in the sense that they will be omitted in very small fonts or at
very poor resolution, because their presence would just clutter
up the characters?

Like PostScript, this system would probably have to generate fonts
dynamically as they are used (although it should certainly be
possible to cache commonly-used fonts).  But it would be a bit
more general by sharing Metafont's "meta-ness".  Unfortunately,
this generality would probably make a real mess out of the
equivalent of the ".tfm" files.  But is this reasonable to expect
somewhere down the line?

-- Richard
   carlson@ernie.Berkeley.EDU
   ...!ucbvax!ernie!carlson