It is not even near to what TeXLive (TeX distribution) provides [LWN.net]

It is not even near to what TeXLive (TeX distribution) provides

Posted Oct 24, 2025 18:52 UTC (Fri) by JMB (guest, #74439)
Parent article: Typst 0.14 released

I was astonished to just overlooked Typst existance.
But it was just one LWN article "https://lwn.net/Articles/1037577/"
and not even any Wikipedia entry at all (AFAIK) ... it is only found
at "https://typst.app/" and "https://github.com/typst/typst/tags".
Even tiny programs have more presence with internet search ...

DTP was interesting when I finish school - and in 2nd semester
at University (as Physicists/Astronomer) in 1992 we all got
emTeX under DOS with a big pile of 3,5" discettes and created lots
of mathematical documents to get acquainted with it.
My old big documents with IBM Writing Assistant was translated
to TeX/LaTeX and since then every letter, book, manual, CD-Cover,
label, etc. was done in TeX/LaTeX - still - so 35 years now.
In 2013 I started making Talks with TeX/LaTex - but the Beamer package
(which had more features) was breaking with the immense amount of
packages I need - so I used Powerdot for that - and stayed with it.

Working with TeX/LaTeX is a mixture of programming, Typesettting and
providing elegance ... and finding errors is only effectively possible
if one can use ... I would call it intuition ... as in most cases I know
what went wrong even when not looking at errors at all but just
from the strange output it provides. Especially if one frequently
creates the document when changing the source.

If someone has a precise notion of how it should look, only TeX/LaTeX is
capable to reproduce it. And the quality is fantastic - zooming in - no problem,
same as zooming out. Tiny but high resolution images - great.
I use several 8k pictures on one slide (coming from Astronomy) is crazy -
when zooming in a tiny picture and suddenly new worlds are visible is fantastic
to see - especially when people are used to Smartphones and Tablets - and thus
such a quality is a shock and a crazy feeling ... similar to material to demonstrate
our brain functions concerning 3D and showing crazy things about optical illusion.

The only thing I hate is to be limited by 4k resolution with 31.5" screen;
8k should have been there since several years ... even Vulkan forces to
optimize for 8k - but no non-glare professional screens with at least 50"
are available. As 31.5 " with 4k does not show the pixel matrix, 4k res
is not really provided ... so higher resolution need a larger surface.
It just demonstrates the lack of quality which is forced by society - and that
all prefer convenience - even if it limits their life quality and also life time.

I liked black boards - and nothing can be seen in the digital regime.
Especially when I think of University Lectures ... on may need 64k
to come near to the resolution. 300 dpi DinA4 created by a printer of 1990-ies
reached 4k - diapositives reached 8k ... so we are stuck with resolution of
the former century when using digital devices.

So I would not be astonished that most people will never touch TeX/LaTeX.
But no real scientist can get around it ... the reason is arXiv, the TeX/LaTeX-Preprint
Archive by Paul H. Gisnparg, which is much to important to avoid it.

And to clearly state it - I have only seen books by Adobe InDesign with PlugIn
MathMagic and a lot of packages for fonts etc. (i.e. one has to pay a lot) coming near
the quality of TeX/LaTeX - thill visibly worst then TeX.

The school books are cruel to see - even some books for University are written
as if no care was used in creating the print ...
So Typst has nothing to do with TeX/LaTeX - which has no rival at all.
But with Donald Knuth as author it must create elegancy and precision.

And a fast workstation can nearly be fast enough to see the result of 50 pages in less
than a minute - even if including a lot of 8k PS images. Of cause it could be faster,
but fast enough to get a high quality job done.

It is similar to Unix - which is said to be user-friendly; but being just picky about its friends!!!
TeX/LaTeX is not easy - it is for sure not in the comfort zone - but for highest quality there is
nothing which can compete.

And to show that it is difficult - even some really good scientists, former colleagues, were not
willing to listen to me explaining how to find the bugs hindering to submit the paper ... but just want
the errors fixed ... it is all a question of personal taste and the ballance between effort and convenience,
expertise and coming along.

to post comments

It is not even near to what TeXLive (TeX distribution) provides

Posted Oct 25, 2025 6:49 UTC (Sat) by burki99 (subscriber, #17149) [Link] (1 responses)

I understand your focus in comparing Typst to (La)TeX. But to me, the difference between the two is way more a reflecting the Markdown vs. *ML (where * stands for SG/HT/X) split. So while the latter need explicit start/end markers to build a proper document tree (throwing errors at you if the tags aren't balanced, though HTML5 came up with rules to handle misnested tags: https://html.spec.whatwg.org/multipage/parsing.html#an-in...), Typst follows Markdown where a document is a strictly linear sequence of characters, where certain patterns (e.g. + or - at the beginning of a line or a pair of characters such as * or _) will transform your sequence into a tree (and where you can't have invalid Markdown, just output that doesn't mark your expectation if you misplaced your special characters).

Markdown-like languages are therefore always easier to start with, but you will hit limitations (either in what is doable or falling back to start and end markers in one way or the others) once certain complexities arise (e.g. if you need to switch between multiple languages in a single document to achieve proper hyphenation).

Luckily there is Pandoc. If the simpler input representation appeals to you, you can always start with the Markdown-like document and convert it to explicit tree-like representation once you really need full control.

It is not even near to what TeXLive (TeX distribution) provides

Posted Oct 25, 2025 8:59 UTC (Sat) by tialaramex (subscriber, #21167) [Link]

> either in what is doable or falling back to start and end markers in one way or the others

The latter situation seems tolerable? Make simple things simple and complex things possible - right? Markdown style is simple, but we don't need to make complex things simple, only possible.

My impression is that the moment you go off piste in Typst you're writing code and calling code, which is back to the "start and end markers" because that's how function definitions work. When TeX was invented that's scary, even most of the mathematicians don't write code, some of the new Computer Scientists back then don't write code, so in TeX although of course programming TeX is how you get tricky things done, ordinary users don't learn to actually code in TeX.

Today every one of our STEM degrees offers programming except Medicine, and the only reason the Medics aren't exposed is that there is literally no room to slot a programming course into their timetable, I don't even know how they sleep - whereas we found room for the other disciplines. So Typst can just go "To do this tricky thing write a function" and the readers go "I can write a function, that works".

Crucially we're not talking about actual Software Engineering. The code in some statistician's Typst document is going to be of the same standard as their R, or their Python, a good engineer would be appalled, but that's fine because it's not supposed to be engineering, the engineering is in Typst itself hopefully. If Typst is a skyscraper, the functions written in someone's Typst journal paper will be like a bookshelf in an apartment in the skyscraper. Non-structural.

It is not even near to what TeXLive (TeX distribution) provides

Posted Oct 25, 2025 15:18 UTC (Sat) by jreiser (subscriber, #11027) [Link] (1 responses)

> The only thing I hate is to be limited by 4k resolution with 31.5" screen; ...

> As 31.5 " with 4k does not show the pixel matrix, ...

Please clarify the resolutions involved. What is the size in pixels of the logical image ("canvas"), and what is the size in pixels of the physical display? The X11 Window System software (common on Unix-like systems for decades, and supported by video cards for almost as long) allows allows a logical canvas of 32767 by 32767 pixels, with today's common viewport (physical display) of 1920x1080 (HDTV), 1920x1200, 2560x1440, 3440x1440 (34-inch "wide screen" by LG, MSI, ASUS), 3840x2160 (QHD), or 5120x2880 (Apple Studio display). The viewport can be positioned at any point in the logical canvas, and it is common to "fly" around the canvas in real time using mouse or arrow keys, thus displaying a view-port sized subset of the canvas. It is unreasonable to require simultaneous display of the entire canvas at pixel resolution. For example, the painting "St. George and the Dragon" by Peter Paul Rubens is large (309 cm × 257 cm) and so detailed that every square inch contains a recognizable sub-image whose physical pixels are the size of the point of a round toothpick. Early "high-resolution" digital scans of the entire painting could not resolve such detail.

It is not even near to what TeXLive (TeX distribution) provides

Posted Oct 27, 2025 8:25 UTC (Mon) by taladar (subscriber, #68407) [Link]

It should also be pointed out that at 32767x32767 pixels (or 2^15 in each direction) you need 1GB per byte you want to store for each pixel. Or in other words at the usual 32bit per pixel you would need 4GB just to store the screen and another 4GB for double buffering.

Even at 8k screen resolutions ( UHD, 7680 × 4320) you are already at roughly 128MB per image.

Now keep in mind that this is just the screen, to make this useful for e.g. 3D you would have to load dozens, hundreds or even thousands of textures of a comparable resolution for a given 3D scene.

Modern RAM and VRAM amounts just can't supply that kind of data fast enough, especially when you want the ridiculous 144Hz some gamers would like at the same time.

It is not even near to what TeXLive (TeX distribution) provides

Posted Oct 29, 2025 14:46 UTC (Wed) by peregrin (guest, #56601) [Link]

> I was astonished to just overlooked Typst existance. But it was just one LWN article "https://lwn.net/Articles/1037577/" and not even any Wikipedia entry at all (AFAIK) ...

Typst actually has a Wikipedia entry: https://zh-yue.wikipedia.org/wiki/Typst

More seriously, a draft entry for Typst exists in the English Wikipedia (see https://en.wikipedia.org/wiki/Draft:Typst) but official submission has been declined multiple times due to insufficient third-party coverage.