What?? I've never met anyone who thinks that. Personally I think it's a very hard standard to beat. I rarely see a document with a non standard font that wouldn't look better in computer modern. Especially for math.
To any professional type designer, it's a terribly balanced font aesthetically, basically every letterform. It looks like a font designed by an engineer, not a designer who knows how to create and balance shapes and lines that are pleasing and easy to the eye. Which, of course, is exactly what it is.
> It looks like a font designed by an engineer, not a designer
This is exactly why I like it. Designed by an engineer for the purpose of typesetting engineering articles and books.
Most of the other alternative fonts feel like they were designed by a designer with all the unnecessary flairs that would make a great title/heading font to pad the portfolio but are needlessly distracting in a paragraph of dense technical text interspersed with equations.
Keep in mind that Google / Facebook / Twitter / Apple / Microsoft / etc hire hordes of professional designers and the result is a horrible user experience for everyone. I much prefer Wikipedia and Hacker News to most "modern" web design by professionals.
I don't know where you're getting "unnecessary flair" from. What unnecessary flair do fonts like Times New Roman or Baskerville or Bodoni have? (Obviously I'm comparing with other similar fonts traditionally used for body text typesetting with equations.)
I'm talking about good sense of aesthetic proportion and balance, not "flair".
And what exactly is the horrible user experience that Google has created with Roboto, or Apple with San Francisco, or Microsoft with Calibri? They're actually quite excellently balanced and pleasing fonts for their purposes.
> Most of the other alternative fonts feel like they were designed by a designer with all the unnecessary flairs that would make a great title/heading font to pad the portfolio but are needlessly distracting in a paragraph of dense technical text interspersed with equations.
Serious alternatives were designed by foundries decades or even centuries ago.
Computer Modern is largely a copy of Monotype Modern, which dates from 1896. Monotype Modern was, I assure you, not designed by a mathematician nor was designed for use in printing math books. However many math books had lately been printed using typefaces similar to Modern, and Knuth liked how it looked.
Every criticism I've seen of Computer Modern is an appeal to authority.
> not a designer who knows how to create and balance shapes and lines that are pleasing and easy to the eye
"Your subjective opinion of the font is wrong, because you're not an expert." If OP and GP (and honestly, most people I know) like the font, how is it not pleasing to the eye? What does it mean for the "aesthetics" to be "_terribly_" balanced?
It's not an appeal to authority, but rather to shared taste.
It's true there's no reference manual for the equations for pleasing shapes, because however our preferences are expressed in our brain, we haven't been able to decode.
But to give one example to your question: the serifs in Computer Modern are thicker than the thin stems of letterforms. That's ugly, full stop. It's not balanced -- the proportions of thicknesses is backwards. The entire function of serifs is to taper and/or finish, never to add weight.
Or another: the loop (lower part) of the double-story lowercase "g" simply extends way too far to the right. It makes the letter feel like it's going to tilt and fall over to the left. It's not balanced, period.
So these are just two examples of terrible balance. Does that hopefully answer your question in a way that isn't an appeal to authority?
Edit: curious why I'm being downvoted for this comment, when I'm just trying to answer the parent comment with actual examples.
> the serifs in Computer Modern are thicker than the thin stems of letterforms.
1. Is this really true? I went to the bookshelf and pulled out three Knuth books, and at least to my eye, the serifs don't look noticeably thicker than the letter stems in the unbalanced way you mentioned. The exact proportions could be found by checking Volume E or generating proofs from the Metafont sources (haven't tried that), but this seems like the well-known problem with the "spindly" Type-1 versions of CM that many people use today, than in Knuth's actual Computer Modern as in his printed books. (See https://tex.stackexchange.com/a/361722/48 for pictures.)
2. Is this any worse in Computer Modern than in Monotype Modern, the typeface that Knuth was trying to reproduce? This is the font used for TAOCP Vol 1 first edition (1968), Vol 2 first edition (1969), Vol 3 first edition (1973) and Vol 1 second edition (1973). Those books were typeset with hot-metal typesetting (on Monotype machines), and in fact when Addison-Wesley approached Knuth in 1962 (when he was in grad school) to write a book, he was excited because he loved the appearance of their books. The publishers' move to phototypesetting could not recapture that look, and digital typesetting was starting to become feasible, so he took up the problem himself: he wanted to reproduce Monotype Modern so he needed TeX to typeset it, and Metafont to specify it (and Computer Modern was the result).
My guess is that the opinion you're expressing results from a combination of both the above:
• Yes, what Knuth was aiming for (Monotype Modern 8A, or the look of math textbooks he used as a student, or mathematical journals of a certain period: https://projecteuclid.org:443/euclid.bams/1183544082 ) was very much an early 20th-century / late 19th-century look, and it seems the fashion in contemporary typography circles (I suspect this started with William Morris in the 19th century already) to look down on that period and all that it entails (like larger spaces between sentences: witness Bringhurst's comments about "In the nineteenth century, which was a dark and inflationary age in typography and type design" etc). So, people au fait with modern typographical fashions don't quite like the associated style (Scotch Roman typefaces, etc), while many mathematicians quite prefer it.
• The poor Type 1 versions of Computer Modern cause the letter stems to appear thin, and even more so on low-resolution devices like monitors, causing the serifs to appear thicker in comparison.
Personally, comparing Monotype Modern (as in the first editions of TAOCP) and Computer Modern (in print, and using "true" CM), I don't think Computer Modern looks worse than the source typeface in terms of having terrible balance or appearing like "a font designed by an engineer"; the issue is probably more the "shared taste" you mentioned: Knuth's target aesthetic was itself different.
The slight unbalance and the not-totally-uniform looks make cmr much more readable. The breaking with the typesetting tradition makes it better, because it is unhindered by such useless flourishes as "must taper" or "must look like any other letter fitting into a uniform box of equal greying when smudged". cmr does this while still looking good and pleasing when viewed in whole paragraphs as opposed to single letters.
And where cmr absolutely excels is the accompanying greek, math symbols and typesetting. Nothing else comes even close in getting it readable, pleasing and uniform.
> The slight unbalance and the not-totally-uniform looks make cmr much more readable.
That's frankly a very idiosyncratic take, and not one many professionals would agree with.
By that logic, toolbars with icons would become easier to use if their icons were all slightly off-center from each other -- which is very much not the case.
Part of reading legibility comes from the shapes of words, not letters by themselves. Generally speaking, we read words -- not letters. When individual letters are off balance or especially non-uniform, they draw attention to themselves rather than making the word coherent as a whole. This, of course, is one of the reasons why kerning pairs are so important.
Also, tapering isn't a "flourish" it's just how roman serifs work. And "uniform box of equal greying when smudged" is neither a flourish nor a rule at all, it appears to be something you've invented.
I actually appreciate this answer and am sorry you're getting downvotes.
That said, the given justifications still seem subjective to me (why should serifs only taper? I actually like the added weight because it seems to emphasize the orientation of the letter. Why should fonts be scrutinized according to the laws of physics?).
I suppose to some degree the difficulty I'm having is what I view as an over-emphasis on objectivity in a discussion about aesthetics. Not that your perspective is wrong per se, but I find it hard to ignore the strong preference among my peers for choosing (and commenting positively on) this font.
CM is really light. As an undergrad I had more trouble reading poor quality photocopies of notes in CM that my own notes not using CM did not suffer from.
In an extreme, one of my professors presented from notes and blew up his notes' display size like crazy, and I could never get a reasonably information-dense printout of those by printing multiple pages (those sizes were so large the fonts were too slim for the printer's resolution). Granted, it is an abuse of good typesetting, but I imagine my predicament would be helped with a font that had proper contrast.
> To any professional type designer, it's a terribly balanced font aesthetically, basically every letterform.
It works pretty well balanced in print on systems it was originally targeted to. The main problem with using other fonts in TeX/LaTeX (or anything in, say, Word) is making it so complex equations lay out reasonably. Admittedly a niche problem, but one one that was the original purpose of TeX - and CM works well for it. There are a small number of other options that don't screw it up.
It doesn't work well in print. It's less legible than more common typefaces in comparable sizes on account of its lightness, for certain printing technologies
It works well in print if tuned to the printing technology, as originally designed. You should be able to adjust it (again, as designed) for the print technology you are using.
Most common typefaces are useless for mathematics, so it's a bit of an oranges and apples comparison anyway.
Wasn't Computer Modern created with input from Herman Zapf, and Matthew Carter among other titans of typography? I don't think Knuth just whipped it up in isolation without any feedback, so this whole "it looks like a font designed by an engineer" seems both incorrect and needlessly arrogant?
I don't think Carter had any input. There was some (minimal) input from Zapf, mostly with respect to the calligraphic capitals. Mostly, Knuth was trying to replicate the Monotype Modern. Some of the fonts like cmr17 are especially bad (the extrapolation of parameters to larger design sizes was not correct). That said, I do think that cmtt is a superior typeface for monospace typesetting.
Do you have any examples of math / engineering texts written with Minion Pro? I can mostly find it used for title fonts, or with a ton of line spacing.
I took a look to see the font, which I agree is beautiful.
But I stayed for the content. Interesting! And what a great presentation! I have so many questions. What happened to this in the intervening decade? Is the idea incorporated into any other plotting packages? Did you purchase this font to use in your talks?
The package vfplot [1] is still available and still under development, it's rather hard to use in that there are lots of parameters to adjust to get decent looking output. There's also an issue in that the "dimension climbing" approach means putting an ellipse at each boundary corner and then as many as you can fit on the line-segment between them, there are lots of ways that this can fail for complex boundaries (coastlines, for example). I think I've fixed this using the fact that "a line segment is a degenerate ellipse", so one can actually calculate a "distance" between a line-segment and an ellipse in almost the same way as one calculates the distance between ellipses. The code for this is in a branch on GitLab [2], but there is still quite a bit of work to do for the 2.0 release (later this year?).
Thanks for the example! I think the scribd pdf viewer has some layout issues, but I can now recall that I've seen this font in quite a few places before.
It doesn't look bad, but I still don't think it quite works for dense paragraphs of text compared to Latin Modern. I do really like how it looks on the slideshow in a sibling comment, though.
Tex Gyre Termes could be. For an example, please look at https://www.texmacs.org/joris/zcomp/zcomp-abs.html (you can download both a pdf and the TeXmacs source---note that TeXmacs is not based on TeX, you need to use the TeXmacs program to edit it comfortably and to obtain a pdf).
Language and glyph support aside, the unbearable thinness of Computer Modern makes it very difficult to read on screen. Maybe it looks better in print, but for computer monitor, Latin Modern is a better alternative IMO.
My understanding is that the "unbearable thinness" is what happens when Computer Modern outlines designed to accommodate ink swell (associated with older printing methods) are used on modern display and raster printing surfaces without adjustment. If you check the typography in TAoCP, which Knuth fine tuned, it doesn't seem so unnaturally thin.
The art of optimizing a typeface for print is pretty elaborate, and if you're genuinely interested in learning how to do this, https://typedrawers.com is a good place to start. It's where all the font engineers, as well as foundry owners, hang out.
No, seriously, when UTCS replaced the 300dpi laser printers with 600dpi models, Allan Emerson was very upset because his papers suddenly looked different. :-)
FYI on modern LaTeX engines like LuaLaTeX (and also XeLaTeX) Latin Modern actually is the default.
But Latin Modern and the default Type 1 version of Computer Modern have exactly the same thickness, so if Computer Modern looks thinner for you then it must be some weird screen effect. (The hinting is a bit different, but I wouldn't have expected that to have a significant effect on current systems where hinting is often ignored anyway.)
AFAIK, Wikipedia's info on the Latin Modern regressions from Blue Sky is accurate. Neither version is the original CM font or has had as much fine-tuning put into it, but (EDIT) the Blue Sky version was at least worked on by paid professionals.
First, the main way PDF readers make strokes thicker is via hinting, so it depends on how Computer Modern was embedded into the PDF for starters.
Second, hinting distorts letterforms so many top-quality renderers (like on macOS) don't use it, preferring more accurate letterforms instead. And Preview on macOS is certainly "decent".
Third, "updating" isn't going to do a thing. Whether fonts are rendered as hinted or not is a design decision taken when the PDF rendered was built from the ground up. No "update" is going to change that.
It seems that modern type designers don't like Modern fonts. Font design, like all design, is a style-based endeavor, where things come into and go out of fashion. Modern fonts are out, currently.
Personally, Computer Modern is a little spindly for my taste (and eyesight). Computer Concrete, on the other hand....
Maybe you and the people you know are very used to computer modern? Complaints I hear often compare it to the fonts (and typesetting in general) of people’s preferred printers from before computer typography, so a particular typeface provided by monotype and set by a particular press.
I think a modern typeface (that is, one with contrast between thin and thick strokes among other things) is more acceptable on computers now that high-dpi displays are more common, but the resolution still isn’t perfect and that style isn’t so fashionable in most cases today.
Personally I dislike the shape of the lower case Roman t (the curl at the bottom is way too long and comes too high) and I dislike the italic u and v (as I can’t tell them apart). I think the serifs on the capital T come too low as well.
The wide array of mathematical symbols in latex don’t always go with the font and they can fail to go together well but that is a separate problem (many fonts won’t include them and the symbols weren’t all designed together.)
> Personally I think it's a very hard standard to beat.
I wonder if Knuth himself would even agree. Open Concrete Mathematics next to a volume of TAOCP and see for yourself. It might not be a huge issue for a paper, but anything longer typeset in CM (like many PhD theses) could benefit from a better font. Even for math, there are pretty good alternatives nowadays.
The font in Concrete Mathematics, interestingly enough, is a reparameterized version of Computer Modern. I'd have to look at the MF sources, but if I recall correctly, it doesn't have any changes to the actual character programs. The "math italic" is the Euler typeface designed by Hermann Zapf. While this was implemented in Metafont, it is not really a "meta" font. Instead, the grad students who worked on this (under the direction of Chuck Bigelow and Kris Holmes, IIRC) instead created MF outlines for each size independently. There is no shared code for different sizes or even for common aspects of the characters in the font.
I used to use LaTeX in my physics days, including my phd thesis.
I recently compiled a document, 20 years later, and find the font horrible.
I was probably used to computer modern and did not realize how awful it was. There were not tons of good fonts easily useable either, that could be another reason.
I remember being at the pool at the hotel during the 1990 TUG conference and Michael Spivak going on a long tirade against the qualities of Computer Modern. There's a reason he commissioned the creation of the MathTime fonts.
I don't think "horrible" is a general opinion, although it is in the end a matter of taste. But you do often hear people who think that the vector derivatives initially made (Knuth's originals were raster) are too light for comfortable reading on a monitor. There have been some recent uploads to CTAN with heavier versions of CM. I remember them as https://www.ctan.org/pkg/newcomputermodern (book weight) and https://www.ctan.org/pkg/mlmodern.
Knuth's original are not really raster fonts, they are vectors but in a format which is not supported in other systems so they have to be mapped to raster fonts before being include e.g. in PDF files. The important difference is that you can customize for which resolution they should be rasterized, so you can get good looking fonts by setting sufficiently high values for your screen and zoom requirements. (Of course this doesn't help if you only get a finished PDF from someone else and for the Type 1 variants it's no longer important anyway.)
People did. I was only talking about the original created by Knuth, the fonts used by default in modern TeX installations are normal Type 1 vector font versions of the font. Most other current variants (like Latin Modern, New Computer Modern, etc. which also include OpenType versions) are derived from these Type 1 fonts.
As contrasted with OpenType? I can't speak for the developers of the fonts but when I compile, Type1 is what I happen to use. I note that the documentation for the MLModern fonts say: OpenType support is planned for a future version.
Yes, as contrasted with otf/ttf. I think of Type-1 as obsolete technology, not often used for new fonts and only supporting 256 glyphs per font. But I guess there must be some upside that I don’t know about.
I want to hate it because it seems so old fashioned, and I constantly search for alternatives, but I keep coming back, because no other font allows me to read dense text for hours at a time.
I have a printed copy of Knuth's "Digital Typography" on my desk (hardcover from 1999). The table of contents does not look too great, but the rest of the book is a piece of art in itself.
it is a lot better. Quite a differemt font, even. I never bother to change the original in latex for my own things, but whenever I have a long paper that I really wat to read typeset in CM i make sure to print it.
\usepackage{mathpazo} is pretty nice. Palatino plus good math fonts.
The problems with CM are twofold. First, it's an inconsistent and amateur remake of Modern, especially italics. The bowls and straight lines are different angles and widths. Second, the postscript converted version used on most platforms is weirdly, bizarrely thin compared to the original Modern, which does not make for a good reading experience.
As you point out, there are two variables here: Is it processed in LaTeX, or processed in something else, such as Word? And is the font Computer Modern, or something else, such as Times New Roman?
I think the renderer has much more to do with the readability of a paper than the font. The combination of LaTeX and Computer Modern is certainly pervasive, but if I had to pick one variable, I would rather read papers using a different font but rendered in LaTeX, rather than a paper using a different renderer to lay out a Computer Modern font. For an example of the latter, take a look at this paper, rendered in Word, with Computer Modern font and some LaTeX-based margins and spacings:
Computer Modern Roman is a shibboleth. When I used to review resumes at Google I'd give extra attention to any resume in CMR; we wanted to hire the kind of nerd who has a resume.tex. I don't like the look of the font myself and long ago switched to Postscript fonts but that's a subjective opinion.
In the true nerd spirit, shouldn't you check the pdf metadata to see if the resume was created by tex. I use TeX (well ConTeXt but not LaTeX, but that is a minor difference) for my resume, but without using CM or LM fonts. With Luatex, you can use any opentype font.
I used to do my resume in TeX (in fact an ancient resume of mine along with the plain TeX macros to format it is on CTAN), but I stopped because in most cases, my resume is getting slurped into some automated system and if it's anything other than a Word document, I end up with a lot of pain.
Default LaTeX fonts give me a nasty uncomfortable feeling in the stomach by reminding me of Math and CS tests. I'm not generally sensitive, so it's somewhat amusing a font can have such emotional impact.
As a typeface aficionado I've spent way too much time building font catalogues such as this in the early 2000s; it's good to see that both quality and quantity of freely available fonts has gone up.
> Default LaTeX fonts give me a nasty uncomfortable feeling
> in the stomach by reminding me of Math and CS tests
Me too. The formulas look so dark. Do you know of any typefaces where the mathematics and physics formulas look lighter, and, at least to some of us, less imposing?
I like the Computer Modern font when printed on paper.
On low DPI screens (e.g. 96 or even 144 DPI), Computer Modern feels too thin and spindly.
Recently, I came across the mlmodern (https://ctan.org/pkg/mlmodern?lang=en) font. It is a "heavier" version of Computer Modern and I use it for all my documents.
If you have a certain printer to use, you can tweak the Metafont mode for the fonts and create your preferred look. The definitions for various printers are in the file modes.mf
For example:
% From {\tt stsmith@ll.mit.edu}, 10 May 93.
% With |fillin=0|, the diagonal of {\tt cmtt10}'s `z' is too thin.
% |blacker=.8| too thin, 2 too thick.
mode_def docutech = %\[ Xerox 8790 or 4045 (600dpi)
Using the rasterized fonts is very much an edge case these days. I'd also note that back when I used to manage this on University systems (in the days before dvips would automatically call mf to generate needed fonts), it was often the case that at lower resolutions (<600dpi), the CM code often ran into errors from the necessary mode_def parameters. Xerox printers were especially troublesome because they used a "write white" strategy for printing. A "write black" printer (e.g., the classic HP LaserJet), marked the page by using a laser to charge the parts of the page that should get toner. A "write white" printer charged the whole page, then used the laser to remove the charge from the parts of the page that should not get toner. On a "write black" printer, a pixel was a little bit bigger than its claimed size. On a "write white" printer", a pixel was a little bit smaller than its claimed size.
I was supporting Xerox 8700 laser printers at the time and the settings for a write white printer inevitably caused errors for many characters until I was generating at least a .600gf file. I have vague recollections of the same issue coming up occasionally even when I was printing to a 1200dpi Compaq.
Thanks for sharing! Those must be exciting times. I remember tweaking the fonts only once for printing my diploma thesis back in the 90s; but contrary to the most opinions, I made the fonts even lighter, because of the ultra white paper, we were obliged to print to. It was a very big file due to the 1200dpi rasterized fonts, but the result was better than any print shop could produce at that time.
Like? Because I've never seen XeLaTeX do something that made me go "wow time to go back to the legacy tex-that-doesn't-understand-unicode->DVI->PDF route".
Legacy would be pdflatex in this case, which generates PDFs directly and works with unicode without issues. However, I would also be interested in cases which xelatex doesn't handle too well.
These are purely my own opinions and not official positions of any groups or organizations I am associated with.
The biggest issues with XeTeX (compared mostly with LuaTeX, some of them also apply for pdfTeX) in my opinion are (roughly ordered for importance)
1. the missing support for microtypesetting features, especially font expansion
2. Missing support for new OpenType features. I consider especially variable fonts to be important, but I know that there are also many users missing color font features.
3. The font selection system which uses completely different lookup paths for font names and filenames leads to extreme confusion for users, especially since it's system dependent.
4. It shares the problem with older engines that many things like colors, underlines, etc. require special invisible nodes which can influence line-breaking in unexpected ways.
5. (From a programmers perspective) adding advanced PDF features under XeTeX tends to be much more complicated because of the DVI based backend which makes it harder to control the PDF file directly.
That being said, I would take XeTeX over pdfTeX any day, but I don't see much reason to use it if LuaTeX is available as an alternative.
The reason I switched from LuaTeX, which I agree is otherwise superior (and...the Lua thing!), is that it has some horrible and well-known bug in its font loading mechanism, where if you load a large number of fonts, it sucks up enormous memory and CPU resources. I have documents that simply grind to a halt if I try to process them with LuaTeX, but compile in seconds using XeLaTeX. I hope that’s fixed now, because I would rather use LuaTeX.
The issue is not so much about many fonts, but about one big font. The first time a font is loaded it has to be analyzed and cached and for some fonts this leads to excessive resource usage. You can avoid that in recent versions by using a HarfBuzz based font-shaper instead of the Lua version: Replace `\setmainfont{Some name}` with `\setmainfont[Renderer=HarfBuzz]{Some name}`.
The first time a bug font is used you might still see a slight delay (~1 second?), but it's orders of magnitude faster and does not require so much resources.
Twenty years ago I did typeset a 440 pages or so book using LaTeX. Sadly the editor wanted a "conventional" look and not the LaTeX one. I ended up using the "Utopia" (it's in the catalogue) font as the main font and sneaked as much "LaTeXism" as I could without getting caught ; )
The editor also had zero familiarity with Linux / LaTeX (it was all QuarkXPress back then) and hence wasn't confident in the digital files produced by LaTeX and the converters (say ps2pdf etc.) so in the end a high quality print copy ended being flashed. So it was still a partly "analog" process even though the computer-to-plate age had already started.
I did the opposite once. In a time before I knew Latex, I had to hand in something typeset in Latex. Instead I used Open Office with the Computer Modern typeface, and handed in my work as PDF, and my teacher never figured it out.
Seems likely. I had an issue with issue one of Serif magazine where I could not move the high resolution scans of some images from my computer to the service bureau that was generating the film for the printers. We ended up have to go old school for that and having the printers photograph those pages and strip the image into the film to make the plates. This was back in the days of uploading files to the service bureau on a 28K modem (maybe I had 56K?) and the files wouldn't fit on the Bernoulli drive I had for sneakernetting large files from one place to another.
Unfortunately, no language coverage is listed for any of the fonts: I generally care about fonts I can use both for Serbian (Cyrillic) and English texts, including simultaneously (programming, maths).
I use roboto light sans serif font for my resume. Are there better alternatives to CM for reading on the screen? I found mlmodern so far to be a good alternative, but would like to know from others as well.
I have only dabbled in it yet every project ends up looking stunning. I'm a corporate data scientist and I'm considering using it for more of my research findings and the like.
You really want to use LuaLaTeX (if you don't do so already) and then you can just load any font on your system using
\usepackage{fontspec}
\setmainfont{Whatever your font's name is}
Especially you don't have to deal with installing old Type 1 fonts which is messy, especially when they are not included in your TeX distribution.
You can also follow the instructions in the catalog which for some fonts show how to use them in pdfLaTeX, but these days pdfLaTeX mostly makes sense if you want to submit to journals which do not accept modern engines and such journals probably don't allow you to change the font anyway.
There are the Tex Gyre fonts, but part of what makes Computer Modern in LaTeX look so good is the typesetting quality.
I believe TeX actually solves a convex optmization problem to find the optimum inter-word spacing for each line while considering every line on the page. That's why you don't get those crazy large spacings in LaTeX that you sometimes see in Word.
If you employ the `microtype` package, latex will also be able to make adjustments to inter-letter spacing. This can improve results by evening out the "weight". The results are subtle, though -- nothing like the huge improvements that result from switching from msword to latex.
Another slick trick is that latex (really, tex) lets you alter the weightings used in the optimization, paragraph by paragraph. And you can alter the weightings used in deciding where to put page breaks. You can also supply hints on hyphenation, for those rare cases in which the engine cannot decide what to do. (Note: the system adjusts hyphenation to the language, so e.g. those long words in German will be handled well.)
The optimization scheme is a large part of what makes tex so good. That, and the understanding of mathematical notation.
Anyone who wants to get insights into the early setup of tex might enjoy reading the following Knuth (1979). It's a very engaging read, for something that's technical. (Generally, Knuth is a great writer.)
References
Knuth, Donald E. “Mathematical Typography.” Bull. Amer. Math. Soc. (N.S.) 1, no. 2 (March 1979): 337–72.
Yes, you click the OTF or TTF link to get the file.
Some of them didn't offer a link but you still can grab them from the CTAN repository. You would need to do navigate more to get the file. You can get the Type1 and AFM/PFM files (old font formats back in the day).You would need to convert those format to TTF/OTF because some software don't have support for Type1 and AFM/PFM files. You can convert them with FontTools (python).
Recently, I was posting a preprint to arXiv. I intentionally removed a journal style, which changed the font to Times New Roman.