After developing the initial prototype you see in the webpage, I've since gone back to the drawing board. I'm working on developing a firmer foundation for issues like:

- How do you interleave content and computation? See: https://arxiv.org/abs/2310.04368

- How do different syntaxes make different document tasks easy, hard, or impossible? See: https://github.com/cognitive-engineering-lab/doclang-benchma...

I still very much believe in the high-level philosophy, but Nota will look very different within ~6 months. In the meantime, the single coolest development in the document language space is Typst, which I encourage you to check out: https://typst.app/

Also: the next version of Nota will be written 99% in Rust :-)

Not sure the language you choose matters as much as making the API usable by a wide audience. Sure if performance is a real issue then rust makes more sense than JS but I’m not sure that’s going to be hugely meaningful in most use cases.

I’ve never been a fan of Latex despite writing some mammoth documents over the years. Latex always felt like a beast for academics not for business. Yet there’s often things I wanted to do consistently in Word etc. that have never been easy.

Styles can easily become a muddle. Having consistent numbering and bulleting is a pain and errors can easily creep in.

Tracking changes becomes a real problem when you get into many revisions and that often always ends up relying on a level of trust between parties to not override the tracking. I think there’s a killer app in just fixing this issue with a product that guarantees that guarantees all changes are properly shown from the start of a process to it being fully approved by all parties. Businesses, lawyers etc would love that stuff. Heck if you sprinkle blockchain in you might even get easy funding but I think it’s more of a basic cryptography thing than a blockchain thing - at least it doesn’t need that level of complexity.

Fully agree with this, and having typeset my masters thesis and later my resume using LaTeX, I think that the “authoring experience” is definitely the place to focus on improving, LaTeX just takes too damn long to get something good.

If you’re interested in the “markup to document publishing” space, you might also be interested in the open-source report publishing tool I’m now working on, Evidence.dev (https://github.com/evidence-dev/evidence).

It’s similarly based on markdown, though uses code fences to execute code, HTML style tags for charts and components, and {…} for JavaScript, i.e.

    ---
    title: Lorem Ipsum
    description: dolor sit amet, consectetur adipiscing elit
    ---

    ```sql petal_vs_sepal
    SELECT
      petal_length,
      sepal_length
    FROM iris_dataset_table
    ORDER BY 1 DESC
    ```

    

    The longest petal in the dataset is {petal_vs_sepal[0].petal_length}.

We’re also aiming for first class output options for desktop, mobile, PDF and image export.

Previous HN discussion:

https://news.ycombinator.com/item?id=28304781 - 91 comments

https://news.ycombinator.com/item?id=35645464 - 97 comments

On the other hand, legal systems have effectively been doing the equivalent of git since basically forever. There have been very few law books written from the ground up. All other law authoring, be it by kings, priests, dictators or parliaments, was I the form of diffs to an existing codebase.

1. An immutable set of standard terms, with variable references.

2. A collection of cover page variables, that modify the standard terms by reference.

3. A structured negotiation workflow, where users "propose changes" to the cover page variables with automatic "diff-ing" (redlining).

It's not a product targeted to software engineers, but has always appealed to me as a way to sneak in some engineering best-practices into the world of lawyering :)

Full disclosure: I'm an employee

[1]: https://commonpaper.com/product/

[0] https://nisus.com/pro

I am not a lawyer so I don't know that anything in the previous paragraph is true; it's just based on a recollection of something I was told once a long time ago.

1. That's an overstatement: For "original works of authorship," copyright happens automatically upon "fixation" in a "tangible medium of expression" (e.g., saving to a file, maybe even just typing). [0] And it doesn't take much "original ... authorship" to qualify for copyright protection.

2. Here's A hypothetical example: Alice drafts a contract from scratch as Version 1 and saves it to a file. It's copyrighted; on these facts, Alice owns the copyright. [0] Then Bob takes Alice's Version 1 and modifies it to create Version 1.1: Bob's "original" contributions to Version 1.1 are themselves protected by copyright, which Bob owns, bu with two caveats:

(a) Bob has no claim to copyright in Alice's Version 1; and

(b) Bob's own contributions to Version 1.1 won't be protected unless one or both of the following is true: (1) Bob had Alice's permission to base his "derivative work" on Alice's Version 1; [1] and/or (2) Bob's use of Version 1 qualified as "fair use" (a complicated question in itself). [2]

NOTES:

[0] https://www.law.cornell.edu/uscode/text/17/101; see also, e.g., https://www.adamsdrafting.com/the-contract-drafter-as-copyri...

[1] https://www.law.cornell.edu/uscode/text/17/103

[2] https://www.law.cornell.edu/uscode/text/17/107

The benefits of working at the proper level of abstraction compound. It enables tech like diffs and git, which then nicely solves a bunch of other problems as well. Using Word completely side-steps all those benefits. Sure, you get a few nice buttons, but that's literally it. You are trapped forever with no way forward.

This feels like actually programming in Word and manually highlighting comments to be green or something. It's a travesty IMO.

On top of that, all contemporary word processors I'm aware of have, of course, versioning with diffs. It is just different than git (or other programmer tools.) Just as you are using your tools of your trade and don't know much about MS Word, lawyers use their tools of their trade and don't know much about git. It's like saying that editing POs is superior to Trados, because for a programmer it is but a professional translator is going to tell you a different story.

(Of course, everybody everywhere should be using LaTeX for fine-looking documents in all circumstances. No argument here ;))

It’s a deeper thing. You can hack Word and related tools for coding and eventually it is acceptable I guess, but it’s starting from the wrong foundation.

This ladder will never reach the moon.

Word’s diffs are not “just different”. they are objectively inferior in many ways. I personally witness daily the travesty of government staff’s handling of information.

Word is a fancy digital typewriter and IMO it’s the wrong abstraction for this day and age and cultural issues are the only thing keeping us back. As always.

Edit: academic papers looking like they were written on a 19th century typewriter.. I don’t get this fascination with style, from scientists of all people. Lay down the info, provide the data. Kerning your fonts properly.. oh my god, I need to cool down. I am a hot headed type of guy, sorry about that.

That being said, that's just my reading of your comment and I could be wrong, which is kind of my argument here. If I'm right lawyers don't care about your notion, they use language for something different than mere information encoding. Therefore they need tools that support their use case. MS Word (or word processors in general) might not be the best tool for that job, but it is good enough. Integrating a well trained ChatGPT into MS Word will help lawyers much more than any structured entry form ever could.

BTW, the LaTeX quip was intended to make light of the idea of separating content and style, which goes way back. Consider TeX' age. Your reaction tells me, you think LaTeX is a styling tool, which in a sense it is, and that's what it is about, which it is not. Hordes of scientists (and type-setting professionals) argue in favor of LaTeX (or other type-setting systems) because you just write the content in plain text. LaTeX takes care of the style. TeX files are also just markup and easily git'able. It does make life easier, but it is not as important as some people make it out to be.

I also know people usually misunderstand me because I am a “programmer” and all I see is “code”. I guess that’s fair enough, but I fully understand legal being of a completely different nature from Rust.

What I also understand is that no matter how long everyone argues about it, the only thing that matters about legal is the text. The font, the styling, etc is all secondary. It might be important, but it’ll never be primary. Unless courts start judging differently based on page margins I guess.

The same goes for science. Publishing “attention is all you need” in an 8bit NES font might not be fashionable, but it does not and cannot detract from the discovery within it. LaTex produces the exact same documents (I know it is configurable but we are going for a certain style) and that’s what this is about. Not how the tools work but that we fundamentally even care about it instead of focusing on the primary issues like correctness, openness, accessibility. I’d like academic papers to be APIs actually.

Again I see the importance of styling and appearance in general. It’s just that we start with that and I think that’s problematic and actively harms our progress.

Also, to conclude, I am nitwit. This is just my take.

Edit: A man can dream, right? If a paper was plaintext I could typeset it last minute in 8bit NES fonts if I’d be so inclined. I hate ya’ll deciding how everything looks and works. I know that’s technically challenging, but to me that’s where the progress is. An academic paper like, say, a jupyter notebook would be awesome, not? Would you give up your fancy type setting? I would!

IMO, I think it really comes down to finding a universal mechanism for diffing and 3-way merging things that aren't plain text (document diffing). I think distributed version control can be universal (at least on a data level), how an application renders a meaningful diff for a specific task is incredibly subjective to the document type and task at hand. My point being that I completely agree that plaintext makes a whole lot of sense for programmers and pretty much nobody else. However, distributed version control does not have to be confined to plaintext, it's just tricky to see when all the version control systems we're familiar with are plaintext ones.

It's actually bad at non-linear stuff, which you will have noticed if you have ever been working with hierarchical formats, especially e.g. xml or nested JSON.

Word is bad for a whole litany of reasons, but the reason it can't be easily versioned (atop the format being a literal Goldberg machine requiring inane transforms to properly) is that it encodes a bunch of non-linear formatting instructions. Sure, we can sort-of reason about this stuff e.g. with a hierarchical css+html+js structure, but without a way to render that I challenge you to be able to simply diff that information. Seeing "bold" or "blue" seems simple enough, as long as you also know to which elements it applies and in what layout. So, suddenly you can't reasonably diff the css file without also difficulty the html.

For programmers, we are used to reducing things by their dimensions into fairly linear spaces, this then helps us reason fairly linearly about changes, but doing this from any other context is challenging. Lawyers e.g. perhaps focus on the relations between various clauses, so linearizing their document flow is not very important to them, at least when there exists methods to diff the general textual content without investing much in how they are doing that.

As programmers we see the similarities to editing a code base and that excites us, however we do have a tendency to go off and write frameworks to parse and simplify these things, without ever actually bothering to learn to apply these things. This is not invaluable, but it's a different focus, which maybe explains why lawyers are not in the habit of using git.

We’re in complete agreement. But you can do this, you just need to provide a “renderer” and a schema that describes how your tree structure should merge or conflict. If you want to test out a weird version control for structured data, my email is in my bio.

Something like Wordstar would be better than MarkDown.

Last time I lost patience with doing Legal stuff in word and evaluated alternatives, I was most optimistic about Asciidoc. Unfortunately the ecosystem was relatively anemic… the strong syntax was limited by the tooling.

Looks like there’s been some improvement, maybe I’ll try again. There’s a nice new homepage at least: https://asciidoc.org/

A lot of successful products have been built in this way. I've seen developers get upset with Apple for making successful products out of just giving a nice UI to a piece of open source tech that does the heavy lifting. Like it's cheating.

This even happened with Microsoft, they had so many false starts and changes in messaging that they killed their own portfolios. I suspect at least that is why they "embraced linux" because it was excellent at web, and web wasn't busy changing every month (it has been, but that's a different story).

Apple introduced Swift but besides new Xcode versions I get the general impression their tooling has been far more stable.

https://www.clio.com/lawyaw/ https://en.wikipedia.org/wiki/Clio_(software_company)

Anyway that's my limited experience having dealt with a bunch of them - no expert.

As an academic, 99% of my time is spent doing two things:

1. Writing statistical computations using a language like R or python.

2. Writing English text.

The most important thing about a document language is that it should prioritize those things. For example, here's why Rmarkdown/Quarto is better than TeX. A TeX document starts:

    \documentclass[12pt,a4paper]{article}
    \usepackage{amsmath}
    \usepackage{amsthm}
    \usepackage{geometry}
    \usepackage{enumitem}
    \setitemize{noitemsep}
    \usepackage{tabularx}
    \usepackage{setspace}
    \newcolumntype{x}{>{\centering\arraybackslash}X}

    \newtheorem{theo}{Theorem}
    \newtheorem{prop}[theo]{Proposition}
    \newtheorem{lemma}{Lemma}

    \usepackage{fontspec,xunicode}
    \defaultfontfeatures{Mapping=tex-text}
    \setsansfont{TeX Gyre Heros}

    ---
    title: "Natural selection in the Health and Retirement Study"
    author: "XXX"
    abstract: |
      I investigate natural selection on polygenic scores
      in the contemporary US, using the Health and Retirement       
      Study. Results
      partially support the economic theory of fertility as
      an explanation for natural selection: among both white 
      and black respondents,
      scores which correlate negatively (positively) with education are
      selected for (against). Selection coefficients are
      larger among low-income
      and unmarried parents, but not among younger parents or those with less 
      education. I also estimate effect sizes corrected for noise in the 
      polygenic scores. 
    date: "September 2023"

  \documentclass{article}
  \title{Natural selection in the Health and Retirement Study}
  \author{XXX}
  \date{\today}
  \begin{document}
  
  \begin{abstract}
      I investigate natural selection on polygenic scores
      in the contemporary US, using the Health and Retirement       
      Study. Results
      partially support the economic theory of fertility as
      an explanation for natural selection: among both white 
      and black respondents,
      scores which correlate negatively (positively) with education are
      selected for (against). Selection coefficients are
      larger among low-income
      and unmarried parents, but not among younger parents or those with less 
      education. I also estimate effect sizes corrected for noise in the 
      polygenic scores.
  \end{abstract}
  ...
  \end{document}

I think I get your point, but in practice that part doesn't really get in the way, and if you are doing the same thing over and over (e.g. for the same publication) it's just a template anyway.

I don't love Tex/Latex, but most of the other markdown comparisons that emphasize "it's simpler" are because they can't do as much. Which is fine until you need some of that capability.

    quarto -> knitr -> markdown -> pandoc -> [tex -> pdf | html]

At the same time, the markdown defaults produce a nice, readable paper. The TeX defaults get you something that reminds you of Rubik's Cube and Duran Duran.

Ofc that was a major downside, something other markdown editors figured out - if you give people buttons that make it easy and you make it easy to learn, they will learn what they need.

Second, my point isn't just about the specific issue, it's that this issue reveals how TeX thinks about the world. It thinks you want to spend your time writing TeX. No, I want to spend my time writing English. Here's another example. This is how you embed an image in quarto - it's just markdown:

    ![Caption](path/to/image.png)

    \begin{figure}[t]
    \includegraphics{path/to/image}
    \centering
    \end{figure}

TeX is extremely powerful and lets you create arbitrary documents. This is the first time I heard of quarto, but apparently it makes a lot of choices for you that you understandably don't really care about.

Instead of developing quarto, one could have simply written a LaTeX class that defines a function like so:

    \newcommand{\image}[2]{\begin{figure}[t]\includegraphics{#2}\caption{#1}\label{#2}\centering\end{figure}}

    \image{caption}{path/to/image}

By the way, developing a LaTeX class is not necessarily hard. It is more or less a file whose name ends in `.cls` with all the commands that you typically put in your preamble. It just needs a header of three lines that define some meta data and also supports options. See here for an example: https://github.com/latex-ninja/colour-theme-changing-class-t...

You put it in the same directory as your main tex file or in the system wide TEXMFHOME or user-specific TEXMHFHOME.

https://www.overleaf.com/learn/latex/Writing_your_own_packag...

How I do it...

I keep a directory called LaTeX inside my home directory. Inside that I keep a file with all my frontmatter, myfrontmatter.sty (technically a package rather than a class), and also my biblatex file and a scan of my signature for signing letters. When I start a new LaTeX document I add the line \usepackage{/home/nanna/LaTeX/myfrontmatter} to the top (note, no .sty). This keeps my frontmatter minimal and tidy.

Inside myfrontmatter.sty:

  \NeedsTeXFormat{LaTeX2e} 
  \ProvidesPackage{/home/nanna/LaTeX/myfrontmatter}[2015/01/01 by me]

  \RequirePackage{amsmath} % Just replace `usepackage` with `RequirePackage`
  \RequirePackage{amsthm}
  ...

  \addbibresource{/home/nanna/LaTeX/biblatex.bib}
  ...

  %% Macros like for inserting my signature
  \newcommand{\mysignature}{\noindent\includegraphics{/home/nanna/LaTeX/signature.png}}
  ...
  \endinput % Not sure if this line does anything?

I exposed this very problem and ChatGPT proposed exactly the same solution, and also another one using a custom environment.

People (as in, the majority of people) will not be comfortable using a tool that is so unintuitive and hard to use that you need to use an AI to help you in writing.

Writing a document is not supposed to be hard and require assistance to do.

Manuscript composition used to be: write your document by hand or with a typewriter, handwrite some notes in the margin, throw in some pages with your figures on them, then let a professional typesetter take care of all of the technical details of making a typeset document for printed output. This was a whole separate career, and the typesetter would sink almost as much time into making your document look pretty as you put into writing it.

If you are using LaTeX, you are taking on the role of the professional typesetter yourself, and you need to make some specific technical choices to get some output from it. This can be a problem if you are inexperienced and don't know which choices to make or in a hurry and don't want to make any choices, but is also good insofar as it lets you actually produce a professional quality document if you have the time and expertise to do so. The difficulty involved is at least an order of magnitude less than doing composition of metal type.

If you are using markdown (or whatever), you are just punting on having a professional document at the end, and/or letting a system make all of the choices for you (often badly), or perhaps expecting to still hand off your document to a professional at the end for proper typesetting.

I am not quite sure whether this is satire?

I'm not saying you should love TeX, but it's a bit like saying you hate assembly language - if you have the wrong abstractions (writing a 3D game or a web page using assembly language) of course the experience will be beyond frustrating. I don't hate assembly language, but I generally don't need to touch it because higher order abstractions generally suffice. If I am optimizing my compiler output, though, then it's a tool I can use.

Ofc if I have the wrong or missing tools while using assembly language, or any other TBH (python, html, etc), that is also a source of considerable frustration. Not sure where the "hatred" comes from, but perhaps you encountered a poorly done package or editor?

If it's so simple, why isn't there QuarTeX that does all of that and removes the extreme verbosity barrier?

And users of LaTeX are probably not knowledgeable enough or too busy to publish their opinionated subset of LaTeX as a class. I don't know for sure. There is no central body that has an interest in removing barriers, so you might as well ask me or yourself why I or you haven't published anything.

I developed something similar to this at my company, because we write lots of LaTeX documents and need shortcuts like this not only for brevity but also so that we achieve some consistency across the entire team. It's only for internal use though and thus not public.

How would this macro exclude the coverage of other use cases? The language primitives would still be there to do the original code

> There is no central body

That's true for plenty of other syntaxes that at the same time aren't like this

Or you don't want it centered. Or you want several images in one figure environment arranged in some way. Or you want to scale the included image to fit only 70% of the page width. Or maybe you want to rotate the included image by 90 degrees. Many more use cases are imaginable.

You could define the macro to cover all of this, but then you might as well just write the original code.

The first (namespaces) contrast speed (brevity) and specificity (verbosity). Namespacing (long names) are great when you don't have a good idea the full scope - putting your library in namespace means you generally never have to be concerned that terms will overlap.

Macros tend to be the opposite - short, in global namespace, with little flexibility. They constrain the possible output by focusing on a certain goal - this is the point of the macro usually, to avoid some common pitfall without being overly verbose.

If you combine the two, you have a problem - either the macro dictates your preference, or many different macros must exist which, usually being brief or in pursuit of similar goals, will result in namespace collisions.

So of course you can define them, but generally this is better suited towards a specific use case or organization.

Or at least that's the idea. I'm not sure some standard approachable macros are a bad thing, as long as you can redefine them later or pick and choose, but I do get the point of "why not", because they don't want to constrain the language itself. The irony is other systems do this for them and tex becomes more an outsider thing and systems like markdown are far more constrained...

On the other hand there is no way to make everyone happy, and I respect the Tex author decision to opt for a less-is-more strategy wrt to offering building blocks only not templates.

You are of course free to build your template library, and make it popular, but be forewarned that likely whatever it is popular at doing (say, writing science articles), will affect what TeX is used primarily for. Which is the "why".

    >    ![Caption](path/to/image.png)

    >    \begin{figure}[t]
    >    \includegraphics{path/to/image}
    >    \centering
    >    \end{figure}

     \includegraphics{path/to/image.png}

Which is also something the Markdown version can't do at all (give fine control over how the image is positioned). You have to use raw HTML plus probably some CSS if you want that.

Exactly, academics usually don't do that - they write the text with appropriate markup, and then put it in the publisher's template and the formatting according to the appropriate standards is done. You can write your own template, but usually you use someone else's, with the big benefit that you can generally move your content to a very different template of a different publisher with minimal or no changes to your actual writing.

Now how would I do that in quarto - what (and how much) would I need to write to ensure that, for example, the captions for all the images and all the references to the images are all formatted in a specific manner? Because for quarto I would need to make my own template specifying the exact formatting and layout, and a quick browse of its documentation didn't lead me to any examples on how I would control that.

In sane environments there is a split between text and formatting, however, the formatting part has to be sufficiently powerful to meet the various requirements, so there is a certain quite high minimum bar to meet there. Latex works because I can rely that I will be able to easily get my markup laid out exactly as required by arbitrary standards, for any markdown-type standards I need some assurance that this will be possible and easy, that I won't need to (for example) go over all my references and do something to them.

For that matter your equivalent is still one line, it's just \includegraphics{path}. The figure environment is just adding extra capabilities.

I agree not everyone needs to do this, but the trade offs you are illustrating are not "X is better than Y" so much as "X is simpler than Y, and can't do as many things"

For you that trade-off makes sense, great. But I wouldn't generalized it to the value of the tool. I know plenty of academics who are quite proficient at Tex, let alone the simpler Latex, and find it lets them generate the content they want easily enough, given it's power.

This isn't just mathematicians either, though most of the people I know using it came to that out of a need to do math typesetting properly. How would you for example generate a mixed language document with both left-to-right and right-to-left languages formatted correctly?

LaTeX's real problem isn't the syntactic load (easily handled with a decent editor) it's the package system. It can be abused to e.g. generate conference posters well, but it's hairy once you get into the details.

The markdown one you get what you get. Maybe that’s fine. If it isn’t you are out of luck.

The latex one requires more of you but gives you much more functionality in return.

Which is better is going to be entirely situational/personal preference.

   \documentclass{article}

   \title{Natural selection in the Health and Retirement Study}
   \author{XXX}
   \date{September 2023}

   \begin{document}

   \maketitle

   \begin{abstract}
   I investigate natural selection on polygenic scores in the contemporary US, using the Health and Retirement Study.
   Results partially support the economic theory of fertility as an explanation for natural selection: among both white and black respondents, scores which correlate negatively (positively) with education are selected for (against).
   Selection coefficients are larger among low-income and unmarried parents, but not among younger parents or those with less education.
   I also estimate effect sizes corrected for noise in the polygenic scores. 
   \end{abstract}

   \end{document}

My latex docs look nothing like that because I put all the boiler plate in a .sty file.

  * Very tight, but very loaded layouts like A0 conference posters
  * Apply a national standard, such as, for example, post-Soviet GOST documentation styling standards
  * All combinatorial explosion of bibliography styling requirements in different international traditions 
  * Make the documents look like a default style in so and so MS Word version
  * Precise positioning of one picture upon another, or text upon a picture for quickfixes in the papers
  * Be able to consciously tweak any of the above

1 - https://simpatico.io/lit Literate markdown reference.

2 - https://simpatico.io/svg svg animation using object to elt scattering and a RAF pump

3 - https://simpatico.io/stree3 A navigable n-arry representation of diverging, deterministic state

  * Foobar
  Foobars are great.

  ** Warning
  Foobars are not to be used with Booms.

  Foobars are great for reading, writing, and flying. This text is outside the Warning subsection.

YMMV with use case, obviously.

It's pretty extensive. I still use it for my own writing, although I'm probably the only one.

I've used it to create operators for stuff like:

* Lead caps

* Formatting dialogue in a screenplay

* Marking something that's a prop in a screenplay

* Parametric links (e.g. defining @john so that it creates a link to the john handle on X/Twitter)

* Tooltips

Your question about content and computation is difficult. When I was writing docs for my side project I would have liked to have done something similar to having an interpreter run in the page itself and have interactive code you can play with. But such an approach wasn't quite practical (ive seen some top-tier docs do this though!.) Though I ended up writing all my code examples in such a way that they're tested in the unittests. So I at least know if anything breaks.

Good luck with the project

While LaTeX is cool, and I use it extensively, I personally feel that it has not adapted quickly to various use cases. It is not _easy_ to compile into different formats for consumption, and sometimes the layout issues are quite hard to debug. Efforts such as these, even if they do not take off, might give the LaTeX community enough to think about what to focus on for improvement...

That's a good idea, would be nice to optimize that instead of sticking to the poor decision of markdown to double asterisks for the more common bold formatting while also wasting another _ symbol in the process

Did you use Nota to write this paper?

I think the big different between quarto and typst is the scope. quarto is a tool for combining prose and computation to generate many different output types (HTML, PDF (via latex), PDF (via typst), PPT, ...) through the power of markdown/pandoc. typst is a typesetting system for turning plain-text markup in to beautiful PDFs.

I think you're much more likely to want to write typst by hand than latex, but out of the box it doesn't provide any tooling for combining writing and computing (if that matters to you).

I haven't really used quarto so take this with a grain of salt, but from what I can see it is much more declarative, you just declare the content of your document and pick a template to show it. It feels simplier, but at the same time what if I need to customize something here and there? Looks like there are extensions that can be programmed, but they are more like second class citizens that you are not suppose to use normally.

You're right that typst is _very good_ at extensions, and likely will always be superior to quarto when it comes to that. The fundamental advantage typst has is that it's a "greenfield" project, and a very well-designed one at that, especially when compared to TeX.

> Looks like there are extensions that can be programmed, but they are more like second class citizens that you are not suppose to use normally.

We take quarto extensibility pretty seriously! "Simple" customization is available without need to program extensions, mostly through metadata configuration and classes and attributes in the document. This covers the basics like CSS, layout, document listings, etc.

For slightly more sophisticated extensions, you can create "filters" that operate directly on the document AST, either using the built-in Lua extension API or reading/writing a JSON representation (these are both built on top of Pandoc's capabilities, which quarto leverages extensively).

For reusable, packageable functionality, the extension system as it exists today is simple but certainly meant to be used "normally". It's how custom formats (the common, concrete use case is to provide different styles for particular academic journals) are defined and used.

I actually don't agree with this. I think _not_ having "essential dynamism" where it's not needed is actually a feature, not a bug.

I couldn't agree more. Dynamism is great for web _apps_, but it's the last thing I want in a document

There are fantastic, beautiful interactive experiences (for lack of a better word) that are obviously not documents (they can't be represented on paper, there is code running) but they're not really applications, either (they are fully offline, self-contained, state that's only evident on the page).

But they are, universally: dynamic.

Examples: - https://ciechanow.ski/mechanical-watch/ - http://worrydream.com/TenBrighterIdeas/ - even : https://neal.fun/space-elevator/

This is what I think the future of textbooks and presentations should be. But I think part of the problem is that not only do we not have tools geared toward creating them, we don't even have a name for these things. If we say "document", they flunk the pdf test. If we say "web application", they are lumped into the same lumbering category as office docs and enterprise software.

Maybe Nota is a step in the right direction. But it'd be an even better step if it didn't call itself "21st century documents", if for no other reason than to defend against the valid criticism you levied against it.

That term is really quaint these days, and it doesn't fully capture what you're talking about. IIRC, it was more prepackaged animations, photos, video, and music; rather than "dynamic code running on paper"

"Dynagraph"? Sorry, that's lame, but there's my entry.

For more examples, see [2].

Very much agree that this is what explanations and presentations should be in the modern age. I think a documentation language (what Nota and Typst aim to be) is still needed in this age of Large Language Models, when the ideas are more complex than those expressible by natural languages.

[1]: https://en.wikipedia.org/wiki/Explorable_explanation

[2]: https://explorabl.es/

I think that one is rather a special case. You could print it out with the first still of each bit and probably get out 90% of the content/context, and modify document to print out a few stills from each interactive part and get the same intellectual content. In it's case, the interactivity is superfluous, it's there to spark joy. It's really an example of a document with outstanding figures/visual aid - the interactivity is just a bonus.

Similarly, the space elevator could be a picture book/pdf. The interactive bits there also spark joy.

The tenbrighterideas page mostly annoyed me. It's just bastardizing structure to be "interactive", which is to say most of the information is hidden away behind a bunch of clicks and it could have been a document, with one page dedicated to each idea.

I'm sure you can convey the same information in text format (like you say, you could just print out the page), but these particular sites would be a lot weaker, because part of their explanatory power is the interactivity.

The original quote was that dynamism was "the last thing I want in a document", and I think these interactive diagrams and explainers directly show how useful dynamism can be in conveying information.

That's not to say that all dynamism is good - I don't usually want you to use Javascript to just load a new page, my browser can do that just fine - but every medium can be abused. That doesn't mean that medium is bad!

To my point it's "the last thing I want in a document", I stand by it. What I mean is I should be able to print it and really lose nothing central to the content. Yes, digital offers features which may enrich the experience/use/navigation, but at the same time there's questions of accessibility and ease of parsing. If I _have_ to interact with things to get the information//content out, it's effectively a web app, and not a document, and if done wrong it actively interferes with my ability to absorb the information. IMO the brighter ideas page firmly falls under that last point.

The watch page is wonderful, and the visuals and interactivity is done masterfully, in such a way it's obtrusive and not _required_ to understand the document.

I'd classify the elevator page as a web app, but there's really nothing keeping it from being a document/children's book.

And I just really, really, did not like the brighter ideas page. I think the content is good, but the execution got in the way instead of adding to the experience.

For example, with the watch page, if we're defining document by printability, it makes a poor document - while you can print it out, what you'll end up with is a document with lots of static pictures and a bunch of (now useless) text referencing how you can move the pictures to see different things. If I wanted a fully printable document, I'd find a different one that was written with the expectation of being printed - maybe a book about watches, or an entirely static page. That will suit the print medium significantly better than this interactive page.

It makes me think a bit of a science museum, in the sense that most science museums will have a lot of text written around that explains all the concepts they want to discuss - this is how a pivot works, that's what a cow's digestive system looks like, here's a description of a space ship or whatever. And you could collect all this text and turn it into a book, and it would be an informative book that you could read and thereby learn something.

But the value of a museum is that it doesn't just have to be text. You can put a pivot into your visitors' hands; you can show food moving between different parts of a cow's digestive system in real time; you can show genuine pieces of real rockets and discuss what journeys they've been on. The medium allows you a huge amount of extra freedom, and a good museum curator will use that freedom - wisely - to produce an experience that allows visitors to get more insight than they would have if they'd just "printed out" the museum's text and read it all.

That's not to say that written text doesn't have its own advantages - you don't need to visit a book every time you want to get information from it, for example! What's key is that by tailoring the content to the medium that we're employing, we can produce a better result than by trying to apply the norms of a different medium. If we'd built our museum like a book, it would have been a bad museum.

I think a similar principle applies to the web - it comes with its own set of tools and features that differentiate it from books or print media. Some of those are fairly subtle - the ability to reference different pages and sites using hyperlinks, for example - but part of that is the interactivity. And not all sites need interactivity at all times, and the best sites use interactivity only when it adds to the experience (just like the best museums - not the ones that surround you with flashing lights and noise just to distract your attention). But the use of interactivity can elevate a simple text far beyond what a print document can do. I think the watch site is a really good example of what happens when you don't see web pages as "just" documents, and rather embrace their unique qualities.

That's why I don't think it's always helpful to make this category distinction between "document" and "web app", particularly when "document" just means "uses the norms of a different form of media", because the whole point of the web is that it a new media form, with its own features and capabilities.

Reputable scientific journals now post videos online alongside their articles. Interactivity is even better for understanding. But permanence is an issue I suppose.

I think accessibility is a major issue, where the text doesn't make as much sense without the interactive bits, and often the text itself isn't substantial enough to be stand alone.

But also, it's starting to cross the line between web app and document. I can print out a pdf and I just lose peripheral QoL benefits like document search. However, if I try to print a web app I usually lose a lot of the content/context.

Edit: as far as supplemental material goes, I'm all for it. People learn differently, so video, audio, web app, whatever are all great supplemental materials, but a good document should be able to stand by itself.

[1] https://willcrichton.net/nota/

A format is needed that encodes information visually and digitally. The digital layer doesn't have to be visible by default, just accessible when needed.

Yeah.

To author's surprise, Adobe's PDF spec supports JavaScript execution[1]. And interactive 3D graphics [2][3]. Not to mention, audio and video [4].

And "Liquid Mode" for responsive-layout PDF documents [5].

Of course, these "features" were considered bugs by the ISO PDF/A spec (archival, i.e. future-proof), so they were all stripped out [6].

The point being: sometimes a document should be a document.

As for science papers: LaTeX is written by humans, for humans. Custom latex commands and packages allow one to write a plaintext document that is as easily read as the paper it generates.

Which is great for accessibility, among other things.

[1] https://helpx.adobe.com/acrobat/using/applying-actions-scrip...

[2] https://www.youtube.com/watch?v=PKfyFt3zT5A

[3] https://www.youtube.com/watch?v=vW5-1LVtd9U

[4] https://helpx.adobe.com/acrobat/using/rich-media.html

[5] https://www.adobe.com/acrobat/hub/what-is-adobe-liquid-mode....

[6] https://en.wikipedia.org/wiki/PDF/A

I think that there's interesting open space for dynamic documents, but you need some good examples of people using it with taste.

One unfortunate problem is that nobody bothered setting the measure for legibility. On my display the text block is far far too wide. Cf. https://en.wikipedia.org/wiki/Line_length (while we're talking about typography, the fonts for body copy and code are mismatched in size in a distracting way)

As far as formulas/notation is concerned, the notation used in this paper is targeted only at experts in theoretical computer science, approximately the level of advanced grad students or above, who also happen to be pretty familiar with Rust and C++. The gimmicky popups are probably not meaningfully helpful for such an audience, and in my opinion don't really make the notation any more accessible to people without the extremely steep prerequisite expertise (e.g. I don't think this paper is going to be at all accessible to the vast majority of working programmers or computer science undergraduate students).

If you really want to make the paper more accessible, it would be better to focus on reducing the reliance on formulas, reducing the amount of jargon involved, and explaining the concepts and techniques using plain English targeted at a broader audience, rather than trying to add extra colors, click targets, or popups. (A research paper may alternately want to just target experts; that can also be fine. Even for experts this paper is pretty dense though.)

Requiring authors to publish two versions to make it accessible, when the motivated reader just needs a little comfort, is too high a bar. Let them write densely for their primary audience (and this pass peer review) and still give affordances for everyone else.

As for width, font, etc, a stylesheet can fix that. I'm assuming they allow stylesheets to format for each venue appropriately.

That can also really mess with screen readers and other accessibility features.

There's a reason PDF/A spec forbids scripts in any PDF documents that are expected to be available in the future.

[1] https://en.wikipedia.org/wiki/PDF/A

If we want documents that work well on all screens, then we use EPUB.

It used to be. But not anymore. Web browser + js are too bloat for any serious documentation usage.

In theory a PDF is a static document that should display the same in any PDF reader.

Every time I can choose between PDF and EPUB, I choose EPUB.

* More in the space of LaTeX than Markdown (but with elements of each).

* Written in JavaScript (so lots of of people can contribute in a language they already know).

* MIT license.

Nice! I don't know that I have an immediate use for it today, but this looks super nifty. If I did want to write something that needed some LaTeX-y features, and wasn't aiming for publication in a place that required it, I'd give Nota a shot. While I think Knuth is basically a demigod, it's not like he descended from on high, gave us TeX, and said "thou shalt never try anything new ever again".

Missing far too many niceties in comparison to modern languages with more guardrails to protect yourself from silly mistakes. The only way I can write Latex is to heavily rely upon \input{} segments to keep isolated blocks in case I break something through a missed escape.

I keep yearning for a modern take, but it feels like we are stuck in a local optimum from which there is no escape. New platform has to fight with the decades of accumulated inertia and packages which exist in Tex.

Latex is "typographically-complete". Markdown and friends are explicitly not. HTML+CSS is. But what latex has is a reasonable enough syntax that a human can write it by hand, unlike HTML+CSS. Moreover, the syntax, though clunky [1] is designed, as much as possible, to not interfere with the content that the human is writing.

For instance, Latex uses curly brackets {} for macro arguments, because they are least used brackets for content. So when you are reading a latex source, you know that () and [] are content, and only {} are ambiguous [2]. Nota, uses a mix of all three brackets for its syntax, causing additional pain for the person reading/writing the source.

The replacement for TeX/latex is never going to a simpler language. It is going to a language just as complex as latex. But it can definitely be cleaned up and sped up compared to latex. IMHO, somebody should write tex from scratch, improve it's syntax but otherwise keep it largely unchanged. Basically, any plain latex source using some of the popular packages should continue to compile and give the same output. That is the only reasonable way out.

[1] A typographically-complete language will never have a non-clunky syntax.

[2] Escaped brackets \{1,2,3 \} are literal curly brackets. Personally, I only use them for mathematical sets and have defined a macro \set, so in my documents {} are 99% not ambiguous.

I could not disagree more. LaTeX syntax is not 'clunky', it's a mess, and has intentionally been engineered right from the start to be clever rather than consistent. And it's not the syntax only, the obvious mess that is LaTeX's surface goes right on, right to the heart ("the guts" as TeXnicians prefer to say) of the machinery, where no concern is dealt with separately, and anything can influence and break everything else.

Hell you don't even get a semblance of sane text (string) processing or decent numerical computation. Yes, you can do it, in the way you could use a toothbrush or wet wipes to paint your house.

> Latex is "typographically-complete"

Yes as long as one is ready to ignore the fact that quite a few simple things are quite difficult to achieve in LaTeX, e.g. keeping lines the same height and keep register instead of jumping around whenever a superscript is encountered.

> The replacement for TeX/latex is never going to a simpler language. It is going to a language just as complex as latex.

The complexity of LaTeX is just in part due to the complexities of typesetting. It is complex because of an endless litany of bad design choices. HTML+CSS+JS gets a lot of flak for being too complex, but they pale in comparison. For example[1]:

In order to use numerical codepoints to write 東京, you can write any of:

    ^^^^6771 ^^^^4eac   
    \char"6771 \char"4EAC   

    \char"6771{} \char"4EAC   
    \char"6771\ \char"4EAC 

Compare this to XML / HTML 東京 which is a much more reasonable syntax, not any harder to write, and uses an explicit end-of-command marker instead of the 'clever' space which is highly problematic.

[1]: https://agiletribe.wordpress.com/2015/04/07/adding-unicode-c...

There’s a simpler way:

  \usepackage[utf8x]{inputenc}
  
  東京

(2) irrespective of whether you want to use numerical references or not, the example shows that apparently the authors of (La)TeX are unable to use sane syntaxes for their stuff. It's just a very bad idea to terminate your variable-length commands with a space when a space in the output could possibly follow. Same with identifiers: only letters are allowed, no underscores, no digits. You then get names like `\fooBarBazVI` instead of \foo_bar_baz_6 which many would prefer. These are all trifles to be sure, but they're legion, so you get a software that seemingly takes Death By a Thousand Papercuts as a positive design maxime.

TeX engine obviously will need to be rewritten completely from scratch for the reasons you suggest.

Amazingly (to me) it seems typst is doing even worse than LaTeX, while starting much later! I'm happy to be told they have succeeded in this area of course.

Latex is very old and has the features; they can focus on accessibility now.

I believe the issue is that the better-than-LaTeX language needs to be not just better, but so much better that all the tooling and extensions for LaTeX are ported to it. Before this, it won’t be better than LaTeX. So it’s a kind of a chicken-and-egg issue.

The correct way to do it is:

Your favorite markup -> pandoc script and includes -> Latex

- https://pandoc.org/

- https://quarto.org/

- https://markdoc.dev/

- https://mdxjs.com/

- https://typst.app/

Its like walking into a historic European city that has architecture going back millennia and arguing for a great new building design. Greenfield space is scarce and people will not just demolish old structures to try something new. They need to sense overwhelming advantage.

The analogy gives some hints as to what needs to happen for a new approach to take hold. In building construction, massively better use of space was one example: For better or worse, use of steel and reinforced concrete opened the vertical dimension and the rest is history.

Is there such an unexplored dimension that could entice people into yet another document format to "improve" on ascii, restructuredText, wikitext, markdown, tex/latex, asciidoc, html etc. etc.?

The stock answer is some sort of semantic hypertext infrastructure. The original vision is still unfulfilled. If we assume that the walled gardens of today are just a bad nightmare that will pass away, in a re-decentralized web one would need modern, user-friendly and empowering document writing infrastructure.

But there might be other dimensions that would elevate document writing and sharing to new heights. The beauty of innovation is that it is not bound by conventional rules and pre-existing wisdom.

You need efficient documents formats, which means formats that don't try to do everything, which means that you have to segregate different document types in different files - e.g. tables in some standard spreadsheet format - in contrast to HTML that tries to do tables, graphic (SVG), etc. on its own.

Separating each thing in their own file lets the user choose which programs they want to view/manipulate documents (rather that the one-size-fits-all browser), help with distribution across the network, and help with low storage/low bandwidth situations (or do more with the hardware we have right now).

So from my perspective, the answer is negative: there's nothing new to invent. Specialize and refine what was already invented.

As mentioned in another comment, rendering to high-quality PDF is an obvious need/question: can it do that?

(Perhaps) better would be rendering to LaTeX for compatibility with existing system.

If that is so, I think someone needs to wrap Nota into a product for it to take off. Because while the results look great, fiddling with node.js to build a document is too much work -- it's like Latex all over again. Most people will prefer to use Notion or a word processor.

LaTeX, or more generally, the TeX family, never went away. It's still the go-to toolkit for writing journal papers in most of the STEM disciplines.

Moreover, Nota is not a direct competitor to the TeX family because Nota generates something for web browsers to read, whereas xxxTeX generates something for PDF readers to read. (And yes, I know xxxTeX can generate other output formats but most people use it to generate PDF.)

Nota feels to me like it could be plugged in to a static site generator. Except then you have to get the two to cooperate. Still, I think that's an achievable hope for someone a little more dedicated than I am.

Have been using AsciiDoc for the past few years and loving it, only falling back to Markdown on places where AsciiDoc is not (yet?) available. GitHub and GitLab, for example, supports rendering AsciiDoc. PyPI unfortunately has not supported it, but more seem to be looking into it [1], which is great.

--- [1] https://github.com/pypa/readme_renderer/issues/205

I thing Emacs org-mode is better and less Javascript-dependant. Also https://docusaurus.io/ is very nice and flexible.

I suggest to look at these two to extend nota in similar way

Another option I saw was Quarto [1]. Maybe even a simple static site blog like Jekyll can be used as well where i just edit the output HTML as needed? What do you all recommend?

[1] https://quarto.org/

[1] https://git.uberspace.net/homepage/

Here's their tutorial on building a blog with Astro: https://docs.astro.build/en/tutorial/0-introduction/

[0]: https://astro.build/

[1]: https://sjer.red/ source at https://github.com/shepherdjerred/shepherdjerred.com

I strongly disagree in the general case. PDFs all the way.

cough docx cough

And yeah, it's a real problem for data portability and preservation. And no Librefoffice isn't up to the task (onlyoffice seems to do better).

- Pandoc Markdown and Pandoc : https://pandoc.org/MANUAL.html

- PagedJS : https://pagedjs.org/

- Make & Python as glue and helpers for compilation

I manage my references using Zotero like any other academic writer. The configuration is less than 100 lines and I can get a pretty solid result using only basic HTML/CSS skills intertwined with Markdown. You sometimes end up with weird formatting issues but there is nothing you can't fix using HTML/CSS/JS. My manuscript has images, figures, tables, code, etc...

It's good to see people trying to tackle the problem of formatting documents again. LaTeX is good but not for everything and the ecosystem is extremely hard to understand. Word, Pages and other similar tools are... proprietary. What would be a game changer for my use case is to see something like Scrivener with more formatting/layout options: https://www.literatureandlatte.com/scrivener/overview

> Existing document tools like LaTeX, Pandoc, Markdown, and Scribble can (for the most part) only generate static web pages.

I feel like Nota is underselling itself here, or at least not properly arguing for why a new language was created. If LaTeX were a perfectly fine document language, then surely we could extend e.g. Pandoc to provide these dynamic features when rendering LaTeX to a web page.

But instead, a new document language was created. Why?

You could write scribble renderer that renders the document as a server that serves the document in parts over Ajax.

1. https://pandoc.org

But I think Nota goes about this a bit too heavy-handed:

    % let nota = @Smallcaps{**Nota**}
    .@Definition[name: "nota", label: nota]{
      #nota is a language for writing documents, like academic papers and blog posts.
    }

Nota: A Document Language for the Browser - https://news.ycombinator.com/item?id=31348316 - May 2022 (55 comments)

> * View documents on any device that has a web browser.

Ah yes, a JS runtime and a browser, 2 things which are feasible to develop and definitely not massive black boxes.

I love the idea.

I just think it would be better being a format unto itself, or at least not requiring JS and/or a browser. Decoupling from these at least permits other language implementations of viewers/editors; browsers are already basically unimplementable by anyone without massive commercial backing.

If the tool aims to allow to create documents and papers and does not support static output then I don't get how that would work.

https://github.com/nota-lang/nota

Though it’s a platform, there’s a runtime and some open substitutes for the bits not open IIRC.

https://github.clerk.garden/nextjournal/clerk-demo/commit/8a...

I think the modern approach might be to recognize that all readers are different, and that they would use an LLM to transform the document into the form that is suitable for a particular reader.

[0]: https://www.gwern.net/index

> 1. npm install --global @nota-lang/nota

That's going to be a showstopper for a lot of people.

[0]: https://idyll-lang.org/docs

Given this is a new project I would really consider using Deno instead of Node. You’ll have a more reliable and consistent developer experience.