Knuth Scotch

Hello, scholars,

I have a twofold question, and every hope that the combined knowledge on this forum can answer it. 

First:  does anyone know which original face was Knuth's inspiration/source for his CR Modern? I find his font very legible and pleasingly inconspicuous -- it knows how to get out of the way! -- but terribly washed out and spindly. As I get older, I find it a real drag to wade through 20 pages set tightly in that font. I was hoping that I could locate his source, and see what the font color of the original was. 

Second: I learned from this helpful comment that his older books used to be set in a face named Monotype Modern 8A (see here for a good sample: https://typedrawers.com/discussion/comment/49777/#Comment_49777. Thank you, Mr Romer!). Does anyone know if this wonderful face has been digitized, by any chance? 

Thank you all for considering my question! 
«1

Comments

  • Hrant Հրանդ Փափազեան Papazian
    edited December 2020
    Sorry, gotta splice this old thing in:

    image
  • konrad ritter
    konrad ritter Posts: 204
    edited December 2020
    Hah! I second that feeling. But, I gotta chase the bitter aftertaste in my mouth with something. It was a great opportunity missed completely. See also: a century of commercial typeface making that failed to produce anything like the original Bembo. Is there hope for mankind?
  • Monotype Modern Extended (https://www.myfonts.com/fonts/mti/monotype-modern) is technically the digital version of Modern 8A, in that it is the digital version of UK Monotype's Modern no. 7, which was their version of Lanston Monotype's Modern 8A. (The US and British companies shared their families with one another, although not necessarily their numbering schemes.) Unfortunately, the digital version also feels much lighter than its old specimens, since the digitizations were made from the master artwork, not printed samples.
  • John Savard
    John Savard Posts: 1,135
    I definitely think that Modern No. 7 from Monotype is one possible source for Computer Modern by Knuth.
    The reason I think this is quite simple: at least in the United Kingdom, this typeface was very much (prior to Monotype coming out with four-line mathematics in 10 point Times Roman) the standard one to use for mathematical typesetting. As is attested to by the famous book "The Printing of Mathematics" by Theodore William Chaundy.
    The only reason I do not quite go as far as to say that it is the source is that it is reasonable to think that American mathematical textbooks, rather than British ones, would have been the source of his inspiration.
    Before Monotype in Britain developed its four-line mathematics system, a similar system, but not quite as good, had been developed by Lanston Monotype in the United States, called the "Patton method". I learned of this from Daniel Rhatigan's papers, which are available online, and which are an excellent resource for this.
    Thus, some Scotch Roman from Lanston is perhaps more likely to be Knuth's inspiration; whatever McGraw-Hill and the other American textbook publishers routinely used.
  • Thank you, Mr Rhatigan. I have the digital Modern you mention, and I'd have never guess it was supposed to be the same as the one I asked about. Disappointing. 
  • John Savard
    John Savard Posts: 1,135
    Looking in the Lanston Monotype specimen book, their Number 9 looks like a possibility.
  • Thank you. Unfortunately, I don't have a link for that specimen book...
  • John Savard
    John Savard Posts: 1,135
    edited December 2020
    Well, I was looking at the wrong specimen book in any case; the Linotype one!
    Instead, a possible match is Lanston Monotype's 8A, and as for a link to the specimen book:

  • Daniel Benjamin Miller
    edited December 2020
    A brief comment, which does not answer the actual original question, but which perhaps tackles the problem in a different way. You comment that Computer Modern is very spindly. You have a few options.

    1. If you're OK with a bitmap font, you can use the lexmarkr option as given in one answer here: https://tex.stackexchange.com/questions/225027/how-to-create-new-font-which-is-thicker-version-of-computer-modern
    2. If you're OK with a Type 3 PostScript font (OK for printing, not great for on-screen display), check out mine here: https://ctan.org/pkg/mpfonts
    3. You can check out New Computer Modern: https://ctan.org/pkg/newcomputermodern

  • It's funny how Knuth loved a certain typeface so much that he wrote a program to recreate it, and yet it's not really that close. It's like that thing where people who don't know how to draw* draw what they think they see rather than they actually see.

    *realistically
  • Knuth –being a CS mind– needed to formalize letterforms, and was [mis]led by Zapf down the false path of expanded skeletons... even though Zapf's own type tellingly doesn't actually fall for that trap. With two huge names involved, it's hard for most people to admit that the result is horrible.
    You can check out New Computer Modern: https://ctan.org/pkg/newcomputermodern
    Is there an easy way to see what the non-Latins look like?
  • Thank you all for your chiming in with your help; I appreciate it. 

    I have to agree with Mr Papazian about Knuth. Recently, I went and read his piece, on 'mathematical typography.' As I scroll down, I watched in horror how he seduced himself down the garden path of his empty formalism. I actually study some of his gobbledygook as part of my job, so I saw what he was getting at. It's all wrong-headed. Why impose the axiom of invariance under rotations on a font shape? It's not rigid-body mechanics, it's type meant to be read at the same angle, horizontally, unrotated. Same goes for Axiom 4, Locality. Only someone who hasn't designed a font himself would ignore that glyphs must work as a whole too -- as a family-- not just at small neighborhoods around a point. Axiom 6 is a dogmatic imposition ("circles are always best!"), not based in any evidence from the history of type. 

    They're all the same, these math-minded types -- easily seduced by their own BS appearance of rigor, as if their little axioms could ever match empirical good sense, the wisdom of the masters, and the consensus of proven experts. If that "Consider a spherical cow" joke was ever a propos, it's in this instance. 

    Lastly, what program must I use to open a Type 3 font file, I wonder? I only have an older version of FLS. 

    Again, thank you all. 


  • Daniel Benjamin Miller
    edited December 2020
    Is there an easy way to see what the non-Latins look like?
    I presume the easiest way is to check out the font files yourself.
    I have to agree with Mr Papazian about Knuth. Recently, I went and read his piece, on 'mathematical typography.' As I scroll down, I watched in horror how he seduced himself down the garden path of his empty formalism. I actually study some of his gobbledygook as part of my job, so I saw what he was getting at. It's all wrong-headed. Why impose the axiom of invariance under rotations on a font shape? It's not rigid-body mechanics, it's type meant to be read at the same angle, horizontally, unrotated. Same goes for Axiom 4, Locality. Only someone who hasn't designed a font himself would ignore that glyphs must work as a whole too -- as a family-- not just at small neighborhoods around a point. Axiom 6 is a dogmatic imposition ("circles are always best!"), not based in any evidence from the history of type.
    It is true that some of the assumptions made by Knuth in METAFONT are somewhat silly. Still, I think he deserves some credit for developing a program which is actually useful for generating variable-size fonts (which are a good thing™). I also don't mind Computer Modern, though perhaps this is TeX Stockholm Syndrome, from reading too many math books. And Linus Romer has shown us with Elemaints that you can finally get a really good-looking font made with METAFONT, which maybe will be finished only or 5 or 6 years after he started it (if we're lucky), and only ~45 years after the creation of METAFONT (with no serious original fonts created in the interim). Besides METAFONT, TeX itself is a great contribution to typography, even though it is perhaps the kludgiest program in common use today (and that's saying something).
    Lastly, what program must I use to open a Type 3 font file, I wonder? I only have an older version of FLS.
    I don't know what "FLS" is here. mpfonts enables the use of these Type 3 fonts in TeX, using a particular dvips command. This produces a PostScript file from a TeX DVI; you can then use ps2pdf to produce a PDF including these Type 3 fonts. Type 3 fonts do not use Bézier curves alone, but instead can use any PostScript code, and so these fonts actually draw the outlines using PostScript (as a direct conversion from METAFONT drawing instructions, instead of an approximation or tracing). Most programs can't use Type 3 fonts directly, but you can look at them and try to figure out what to do if you have the right time and technical knowledge. If you really want, you can see the curves' PostScript code using a text editor.

  • John Savard
    John Savard Posts: 1,135
    Knuth –being a CS mind– needed to formalize letterforms, and was [mis]led by Zapf down the false path of expanded skeletons... even though Zapf's own type tellingly doesn't actually fall for that trap. With two huge names involved, it's hard for most people to admit that the result is horrible.

    Maybe not too many people have come out and said that fonts created with METAFONT tend to look awful, but the fact that Linux, for example, uses TrueType fonts, rather than fonts based on METAFONT, that although fonts can be converted from METAFONT to TrueType, it's not very common for METAFONT to be used as a font design tool...
    that seems to indicate to me that whatever people have had the courage to say, the same end result is achieved by people's actions. METAFONT may be a failure, but its failure didn't cause serious damage to font design, so its flaws must have been recognized in time.
    At least Donald Knuth's TeX remains an important contribution to mathematical typography. It has its quirks too, but it also has no equal.
  • Thank you all for the latest round of comments. In case I didn't make that clear: I have no problem whatsoever with LaTeX, the program. I know it's unparalleled for people who write math and theoretical physics. I was just complaining about the font, CM Roman or whatever it's called. Especially when he could have made something like this. 


  • And, another thing that just hit me. Knuth in his famous essay set out to survey over a century of hot-metal printed Transactions of the AMS, and extolled the volume for 1922. He complained about all the typesetting after that year, and went on to make a font that looks nothing like the 1922 type he praised. (Other than, they're both scotch romans.) Then what was the point? 
  • @konrad ritter If you are referring to "Chapter 2: Mathematical Typography" from "Digital Typography" by Donald Knuth: The article does not say that the typeface of 1922 was the best but rather:
    The exception was volume 23 in 1922 (Figure 1c), which in my opinion has the most pleasing appearance of all the Transactions volumes.
    When you read "Chapter 1: Digital Typography" from "Digital Typography" by Donald Knuth you will find:
    My publishers found that it was too expensive in 1976 to produce a book the way it had been done in 1969. Moreover, the style of type that had been used in the original books was not available on photo-optical typesetting machines. [...] But the results were very disappointing.
    In my opinion "Chapter 1: Digital Typography" indicates that Donald Knuth tried to imitate Monotype Modern 8A, which was used in his older books, and his resulat is called "Computer Modern".

    @Simon Cozens I am not sure if you are considering the historical context: Yes, DVI seems no longer to be a good idea when we have alternatives like PDF but in 1982, when DVI was implemented, PostScript has not been published (not to speak of PDF). Yes, METAFONT has many flaws but Donald Knuth and John D. Hobby developped (and published) some pioneering algorithms along with TeX and METAFONT like fast pixel-filling for Bézier outlines or Bézier interpolation through given points before PostScript was published. To me it seems that the developments in the TeX-World and in the PostScript-World were quite parallel. In "Chapter 1: Digital Typography" from "Digital Typography" by Donald Knuth it even says:
    I thought it would be easy to find mathematical formulas to describe the shapes of the letters. I had seen John Warnock doing similar things at Xerox’s Palo Alto Research Center, so I asked if I could use Xerox’s lab facilities to create my fonts.
  • If 8A was the face he was looking to imitate, then the less said about his attempt, the better. His spidery little font looks nothing like the alleged inspiration. Not a good look for Knuth. 




  • On the one hand, his father was a printer, so it may be assumed that Mr Knuth was aware of the effects of rendering process on type form, but on the other, as he once remarked (not about CM) "Beware of bugs in the above code; I have only proved it correct, not tried it." In that respect, CM may only have been as correct as its skeletal structure allowed, and this during a time of technological paradigm change, when it was not yet apparent how digital rendering would work out. Redoing CM as outlines was no less a transition than doing a digital Jenson.

    When I did my Modern revival, I worked from an old letterpress printed sample. But that was in 2005, a good decade or more after Robert Slimbach (Adobe Garamond), Martin Majoor (Scala) and Fred Smeijers (Quadraat) had addressed the emaciating effect of digital typography on legacy serif type styles, by sturdy compensation for the lack of press gain. There is a trick to the (Scotch) Modern, it has always operated at the threshold of rendering ability.


    Indeed. And I would recommend that anyone who thinks CM wasn't designed with the printers in mind to find a copy of one of the early books set in TeX, such as one of Knuth's from the era, and look at how the font looks in print.
  • konrad ritter
    konrad ritter Posts: 204
    edited December 2020
    Mr Shinn, that's very enlightening. However, I never quite felt that Quadraat solved the problem you point to. It too could use a bit more flesh on its bones, it seems to me. I love that face (back in grad school, I was a Quadraat freak), but again, it could use a Book/Medium addition. 

    Also, for a very lovely and true-to-the-spirit revisitation of a great Scotch, I must praise Mr Kutilek's digitization of Frutiger's Mergenthaler Antiqua, Light and Regular. Which he gave to the world as a free gift. No italics yet, but we are free to hope... 
  • John Savard
    John Savard Posts: 1,135
    edited December 2020
    First, they fall into the trap of seeing the new discipline through the lens of their area of expertise. Knuth says explicitly that this happened to him - he saw rasterised fonts as a matrix of ones and zeros, and so reduced the problem of typography to one of mathematics.
    I would like to defend Donald Knuth a bit here.
    The problem of going from a description of a character in vector and curve form to the ones and zeroes of a raster, what a renderer does for METAFONT or TrueType, is just mathematics. And I think that's all he meant when he wrote what you're referring to.
    Where those curves are to be placed, how they are to be drawn, obviously is not, but is instead a matter of aesthetics and of things like legibility and readability, which, while less subjective, aren't the province of mathematics. And I don't think Donald Knuth ever claimed otherwise. Did he try to come up with an equation for making an S look like an S, for example?
    One certainly can criticize Computer Modern; one would expect that the first design by a rank amateur would have faults. He was doing things himself in order to blaze a trail, as no one else was doing it for him. If TeX and METAFONT would succeed, then no doubt he always expected that people who were professionals at designing typefaces would bring their expertise to the realm he created which was intended to make the use of type easier for mathematicians.
  • Simon Cozens
    Simon Cozens Posts: 752
    edited December 2020
    Where those curves are to be placed, how they are to be drawn, obviously is not, but is instead a matter of aesthetics and of things like legibility and readability, which, while less subjective, aren't the province of mathematics. And I don't think Donald Knuth ever claimed otherwise. Did he try to come up with an equation for making an S look like an S, for example?



     "After learning how to draw an S with mathematical precision, I found that the same ideas apply to many other symbols needed in a complete system of fonts for mathematics. In fact, all of the characters in Figure 13 use the same METAFONT subroutine that I first developed for the letter S... Without the theory developed in this paper, I would either have had to abandon my goal of defining books in a mathematical way or I would have had to stop using all of these characters." (Emphasis mine)

    (It turns out his The Letter S article is also online. You can see that he took the Roman du Roi stuff seriously, and thought that the reason that real types differ from the mathematical precision of the Roman du Roi is not adjustments for legibility, but that the mathematics weren't good enough and needed improving.)
  • konrad ritter
    konrad ritter Posts: 204
    edited December 2020
    There is a sense in which mathematics is involved in type design, but it's trivial, hence unimportant: whenever you transfer anything from material/analog to digital, a piece of math is involved: the combinatorics of binary numbers, and the discrete math of computer code. But, in that sense, math is involved in anything computer-based. So, font design is no different. Also, math is involved in software that relies on Bezier curves. But, font design is not special in that respect. Drawing dinosaurs or digital versions of Duerer's engravings also relies on the mathematics of splines, which Bezier curves presuppose. So, again, trivial. 

    Both the Romain du Roi and the Bodoni/Didot approach to text fonts (not display!) is rationalism run amok. The thought that a few simple dogmas (straight lines and circles only! vertical stress throughout!) can do better than five centuries of accumulated typographic wisdom about how to pass from handwritten calligraphy to legible, transparent movable type. For a guy who grew upstairs from a printing shop, Knuth seems to have missed that basic point. 
  • John Hudson
    John Hudson Posts: 3,229
    edited December 2020
    Mathematical relationships become much more important in developing type when interpolation or extrapolation is involved in either the design or rendering processes. It is fine when making discrete static typefaces to judge everything by eye, but as soon as interpolation is involved one needs to become more systematically consistent and consistently systematic. The relationship of extrema and curves or mark positions may look okay to the eye in individual masters, but in order to look equally good across the design space they need to have the same kind of relationship across the masters.
  • Mathematical structure becomes much more important in developing type for scientific typesetting, such as Computer Modern. I have yet to see a font with extensive support for arrows, operators, and other technical symbols that does not base these glyphs on a strict skeleton.



    Should the arrows pictured above be optically corrected, or should they have the same (mathematical) length? A strict mathematical system might not look as good, but it is more versatile (e.g. it allows to align formulas neatly across multiple lines).

    Often these symbols are deliberately designed to contrast with the “normal” text glyphs. Notable exceptions are Garamond Math and Libertinus Math which do not emphasize the organic/mechanical divide most other math fonts employ:



    All of the fonts above use a mathematical structure for the technical symbols, non use such a structure for the text glyphs. This is where Computer Modern clashes; its design is mechanical thorough-and-thorough. For Knuth, this approach paid off twice: harmony between the text and math glyphs, and – as @John Hudson notes – the possibility of an expansive font family with variable width, weight, and optical size; all of which are important for professional typesetting of mathematical formulas.
  • Hin-Tak Leung
    Hin-Tak Leung Posts: 363
    edited January 2021
    Type 3 fonts - I remember more than two decades ago I have converted some such to type 1. Type 1 any font editor can open.

    I don't remember the exact details, but, ghostscript has a font sampling utility bundled and ghostscript can certainly process / view type 3 fonts, and also I believe that however I converted type 3 to type 1, ghostscript is involved. So under ghostscript's utility directory there are likely two tools, one for font conversion from type 3 to type 1, and another for generating font sample sheets from any fonts. (the generated sample sheets are postscript obviously, but you can use ghostscript's ps2pdf script to convert to pdf...)

    EDIT: the converter program is called wftopfa.ps, the viewer/sampler program is called prfont.ps . The latter is still shipped with current ghostscript, but the former seems to be gone (I assume it is just routine tidy-up - few people needed to convert type 3 to type 1 these days...) - but you can probably get it from older source archive.
  • Thank you! 
  • Daniel Benjamin Miller
    edited January 2021
    For those who find standard PostScript Computer Modern spindly, here is a gift, for TeX use: https://ctan.org/pkg/mlmodern For non-TeX things, if you need OTFs, you can find them here, though they are lacking some of the corrections and TeX-related tweaks I have now made: https://github.com/jagd/fakebold/tree/master/otf