Hacker News new | past | comments | ask | show | jobs | submit login
The TeX tuneup of 2021 [pdf] (tug.org)
120 points by EvgeniyZh 35 days ago | hide | past | favorite | 39 comments

Highlight from the article.

""" And my fondest recollection from that day was the beaming face of A–W’s cofounder, Mel Cummings, as he held those five volumes in his hands with obvious pride and satisfaction. He had spent his life in the printing industry, and devoted it to producing technical books of the finest quality; so I was delighted to see his delight. """

I met Professor Knuth on his visit to Oxford, UK in 2000, and he was gracious enough to sign my TAoCP volumes.

The tug page has the boxed set mentioned, but it does not appear to be the 35th Jubilee Edition => http://tug.org/books/#dek

Nor does the publisher have any joy yet => https://www.informit.com/search/index.aspx?query=knuth

crosses arms, taps toe

Knuth has a way of adding life to what might otherwise be dry material. At least to those of us who admire the man and his contributions, this essay reads like a thriller. I knew from the start that what he said would not matter to me, but yet I was drawn from sentence to sentence, powerless to turn away.

Is it just that I'm a fan, having read [1] as a science student? I don't think so, because when I recommend that essay to a student who is frustrated with msword, I almost always get a report about how intriguing it is.

Beyond the material, I think a big factor is that it's simply intriguing to hear what a clear thinker spends time thinking about.

1. Knuth, Donald E. “Mathematical Typography.” Bull. Amer. Math. Soc. (N.S.) 1, no. 2 (March 1979): 337–72 https://projecteuclid.org:443/euclid.bams/1183544082

I'm a registered user of a weather prediction suite of software maintained by someone from the NWS SOO STRC. I'm not a meteorologist and have only once barely got the thing to work, but I stay subscribed because the release announcements are pure fun.

TAoCP is a great book simply in terms of books. It is a plausible candidate as the most important book from the current era. [1] By extension, Knuth is a great writer and his skill in language goes beyond English into mathematics and the logical notation of computer algorithms...like Tolkien, Knuth invents languages in order to enrich his stories. Unlike Tolkien his languages are useful to other scholars.

[1] Note I said "plausible." The standard for a rebuttal is implausibility.

Another possible line of rebuttal might hinge on defining the appropriate era ;-)

There's only one current era. TAoCP is a work in progress as we write.

Timelords are inclined to disagree.

And to be clear, the objection is quite tongue in cheek.

Yeah I know.

Friction at other's expense is sort of played on the internet.

At least for me...

It's not just the cool kids arguing for the sake of arguing any more.

I mean I can carefully say "I believe X" and someone will aggressively argue that it is false because they don't believe X.

And I know that's not what you did here. My apologies for being a wet rag.

My initial intent was actually a subtle nod to your efforts at avoiding tenditious, tedious argument. Because I know that feel.

My apologies for the failed levity.

Thanks. I know it wasn't ill willed.

Reading Knuth is always enjoyable, no matter what he writes about. There's something about his writing style that always brings me joy, even when he's writing about stuff I don't really care about.

Usage of (La)TeX has exploded in recent years, thanks in no small part to Overleaf. Unfortunately the project itself has still been struggling financially, so if you've got the means and want to support its development then do consider joining TUG, the TeX Users Group, the global org which coordinates it's development and community.

Perks include being posted copies of TUGboat, the excellent magazine that this Knuth piece is a preprint from. They also send you a DVD of the latest TeXlive, if you want it, you get to vote TUGs elections (or stand if you want!), invites to the annual conference, and get discounts on various (La)TeX, typography, and Knuth books.


Some of my collaborators use Overleaf with the git bridge to interact with our version-controlled drafts. The git bridge is broken, and overwrites git filemodes, erasing executable bits without the user's request; even on files that the Overleaf user doesn't touch. This breaks scripts that we need to produce figures. Every time we have a new paper to work on we have a problem.

It drives me crazy. Overleaf claims that this would require a nontrivial overhaul of their software, which I find hard to believe. But their code is complicated enough that I can't just look at it and find the fix.


Overleaf also doesn't work with symlinks properly.

Knuth added Oxford commas to the master web files. So, I guess we can all say that we are done here. Goodnight.

Ironically the link to Knuth's website found in the References [1] is broken due to a linebreak in the url.

It's plain text. Chrome infers a wrong link, a couple other viewers I checked do not infer any link.

The stackexchange username 潇洒张, listed in the references, cannot be highlighted. Does TeX not render chinese characters as text?

TeX is limited to 127(?) characters, there are several font encoding schemes, none of which are suitable for Chinese.

XeTeX and LuaTeX use Unicode, and thus could render the Chinese name as text. But they are extensions of TeX, not TeX proper, which is what Knuth himself uses.

256; used to be 128 way back when.

Not sure about TeX but would be surprised if there were issues with Chinese characters in LuaLaTeX or XeLaTeX. It's very uncommon to write in plain TeX today, Knuth is an exception.

In this document, it's rendered in bitmap.



This is a good point, the title is wrong!

Knuth named it TeX, with an 'out if kilter' e, to distinguish it from TEX, the Text EXecutive programming language developed by Honeywell. (Knuth, The TeXbook, p. 1)


Reading https://projecteuclid.org/euclid.bams/1183544082 and https://www.saildart.org/TEXDR.AFT%5B1,DEK%5D1, Knuth originally named it TEX and later renamed it to TeX.

This ancient implementation of TeX is still widely used, but note that there were historically many other implementations, and the current one is LuaTeX, together with its experimental fork called LuaMetaTex:



Similarly, the (less) ancient LaTeX is still widely used, in spite of the existence of the (not so new, but still actively developed) ConTeXt:


You are confusing some things. LaTeX is a set of macros written in TeX, as is ConTeXt. TeX, pdfTeX, XeTeX, and LuaTeX, are all TeX compilers. LuaTeX is not "the current one" in any meaningful way. The most widely used one is pdfTeX.

I don't see what things am I confusing. But I should have been more clear: when I said that LuaTeX is current, I was referring to how it's still being actively developed. AFAIK pdfTeX and XeTeX get just the occasional bug fix.

That's not a signal of being used at all.

PDFlatex is the most widely used (La)TeX system.

You newcomers confuse "new" as "widely used" and that's not the case at all.

Unix utilities are old as hell and they are used daily, far more than exa and fd.

You misread me, I specifically referred to the older variants as widely used, in contrast to the (somewhat) newer, more developed ones.

No. This is the canonical implementation written and maintained by Knuth himself.

Now there are a few alternative implementations. Wikipedia list:

* https://en.wikipedia.org/wiki/LuaTeX

* https://en.wikipedia.org/wiki/PdfTeX

* https://en.wikipedia.org/wiki/XeTeX

I think all of them have more features like better Unicode support, but probably also more bugs.

For various languages - at least Farsi, Arabic and Hebrew - the common wisdom and vast majority of users (AFAICT) have moved or are moving to XeTeX as the compiler. The polyglossia package, which is the modern replacement of babel, requires XeTeX. IIANM, the effective Unicode + multi-language support of the other two is not at the same level.

And plenty of people, like me, have moved to XeLaTeX because we prefer to type Greek or accented letters, and even math symbols, directly in our source rather than use the TeX commands for these things. Also, we want to use Opentype fonts and Fontspec.

I don't actually think LuaTeX is capable of even handling right to left languages like Arabic, Hebrew and Farsi. Only XeLaTeX is. Or am I wrong?

I think LuaTeX may have been sort-of capable for a while (with enough effort/workarounds by the user), but in 2019 or earlier, there was some work (initially by Khaled Hosny, who was also one of the major XeTeX maintainers a while ago) to have LuaTeX use HarfBuzz. (See http://tug.org/TUGboat/tb40-1/tb124hosny-harfbuzz.pdf) So (after some more work by the LuaTeX team) there's a LuaTeX variant called LuaHbTeX, which is the program invoked (since TeX Live 2020) when you write `lualatex`. The end result is that, to the best of my knowledge, LuaTeX can handle many of these things just as well as XeTeX. (See http://tug.org/TUGboat/tb41-1/tb127fischer-bangla.pdf for an article from 2020, focusing on the case of the Bangla script, which is not right-to-left, but like all Indic/Brahmic scripts has some other issues related to complex text layout.)

Looks like the engine is getting there - but as of 2020 such use seems to still be semi-experimental. But good to known, thanks.

LuaTeX is in no way the "current implementation", it is one of the modern extensions of TeX, but the core is still the same. By the way, Knuth wrote TeX in a dialect of Pascal, and the original code has been automatically translated to C and used in all modern versions.

Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact