14 Aug 2004
(updated 15 Aug 2004 at 01:32 UTC) »
: IKARUS and Hobby's method both combine, from a user p.o.v., on-curve points with circle segments. I have long believed that this is close to the optimal curve construction technique. Hobby gets bonus points for doing it within the ubiquitous cubic Bézier framework!
I'm not sure what you mean when you say IKARUS uses Hermite splines; as far as I know it uses circle segments between on-curve knots. This is confirmed by the FORTRAN code that Dr Karow included as an appendix to his book Digital Formats for Typefaces.
You're right about CMR versus ink spread/squash. Some offset printed books set in CMR were disastrously thin in their setting. Wolfram's The Mathematica Book was one example. Beautifully printed, with almost no ink spread or dot gain, but the CMR type used throughout was most unpleasantly thin! Knuth was somewhat savvy to this issue as the models he looked up to were letterpress works such as early editions of his TAOCP (which are masterpieces of technical setting). He did include some compensatory parameters in his CMR metafonts to deal with print machinery characteristics, but perhaps some high resolution fonts were generated with unwise or default values.
Many digitisations and revivals, particulary Baskerville, Century and similar, have suffered from an underestimation and/or underappreciation of letterpress effects. I believe this issue has been studied pretty thoroughly by revival-makers (revivalists?) in the last 20 years or so.
California sees the light: Terrific paper on "Exploring Open Source alternatives" from the California Performance Review. A state that bans electronic "voting" and is also open to Open Source! What a place.
[Advogato bug: re-editing articles duplicates <p> tags between paras...]