This is unfortunately a very weak point of Typst. You can typeset your text so it looks like a proper acm or ieee conference paper but, for what feels like very bureaucratic reasons, you still need to convert it to Latex for publishing.
If you look at the process for one of the major mathematical publishers like AMS, ACM, Elsevier, or Springer, there is a lot going on beyond just “posting the pdf”. Take https://www.sciencedirect.com/science/article/pii/S0022314X21001980 for example and notice a few details:
- the PDF itself is available and the PDFs are all consistently styled in the journal. They don’t just take what the author throws at them with all their weird fonts and weird spacing - they run them through a standard process to produce something that looks like all the other articles they publish. This depends upon being able to apply consistent styling, and LaTeX is the only system that really does that.
- the title in the article I pointed at has mathematics in the title that is correctly rendered in the HTML page.
- They report full metadata to crossref when they register the DOI. You can see what they report here and you can see that the title is formatted as MATHML. They also report the bibliographic references in a structured format so that others can perform bibliometric analysis on their journal.
- They implement the crossmark protocol so that subsequent corrections or retractions can be found from the PDF.
- They produce a full HTML version of the article as well as PDF. This is important so that it can be read on phones, and it is not trivial to produce a good HTML version that will render mathematics correctly.
- Their submission pipeline supports ORCID authentication so that authors can be properly authenticated against their ORCID. They will also automatically push their metadata for the article back into the ORCID records of the authors.
- They maintain records of who cites the articles they publish.
They have a fairly elaborate pipeline for publishing that attends to such details. It all depends on the authors submitting their article in a specific format so that the downstream processing can handle it. It may seem like bureaucracy to you, but this is how academic publishing works. There is an increasing amount of metadata being collected, and the academic publishing world depends on this metadata for data analytics to establish reputations and rankings.
I’m curious about that first point:
Is this while allowing its authors to just use whatever LaTeX style they want, or with authors using a specific template? Because if the latter, I don’t see how this isn’t just solved with Springer making or approving a Typst template, and then scanning any submitted document to ensure it’s using that template.
I don’t see how Typst isn’t already beating LaTeX at these with it already thinking of HTML as a native output format, while LaTeX is having to bolt things on. Typst might be too young in that it’s MathML export support only got added in v0.15, but its HTML support is getting stronger, and notably by default having much better compiler times than LaTeX (unless you use very specific variants of LaTeX, like PreTeXt.
As for references, Typst already supports .bibs like LaTeX, and supports much richer data structures (as it’s also designed around being a full-on Turning complete language).
I don’t know about the crossmark principle, so you know more there. Perhaps that should be a feature request if it isn’t already.
I also can’t say much for ORCID, but I don’t see how that’s dependent on their LaTeX pipeline.
Lastly, I don’t see how keeping records of who cites them is dependent on LaTeX.
Granted, some of these last things are probably just more intricate than I imagine.
None of these are impressive in any way, nor seem to justify even a fraction of the huge amounts of money they get for what is, no matter how much people want to make it sound sophisticated, upload a pdf with the work of other people, reviewed by other people.
I could literally ask AI to make some typst-supporting slop that in some hours would probably be better than 99% of the garbage tools that fill the academics landscape, having around one hundred additional features and improvements. It is just a money machine where the fuel is fame. The only thing that matters is to be a well known name. Typst is slow to adopt because for the moment it is not needed to get money.
I completely agree that none of what they do justifies the amount they charge to authors for making their articles open access. That’s why I spent two years of my life building an open access pipeline for a scholarly society - so we could reduce the charges to zero. We will probably be dropping springer as our publisher as a result. I now use claude code so it could be cut back to six months, but it’s still a pretty big project.
Typst has been slow to be adopted in part because so few people have requested it. We’ve published papers for 11,500 authors over the years and only two people have asked for typst so far. I still think it would be cool to support it, but it’s going to be a lot of work to build it into our pipeline and so far I keep finding more important things to work on.
The publishers all have their own document classes (Springer, Elsevier, AMS, ACM, IEEE, etc). They certainly check that the document was compiled with their template, but there is a lot more involved because of the tomfoolery that authors engage in like typesetting their section headings in comic sans or changing the bibliography style or modifying the \maketitle that generates the top matter. You’d be surprised some of the nonsense that authors deliberately stick into their articles because they like it better than what the journal uses. It often takes a lot of work to unwind the crap that authors stick into their articles. The publishers typically compile the articles inside a sandbox to make sure that it can’t incorporate strange fonts or packages that are inconsistent with the style of the journal. It sometimes becomes an adversarial situation between authors and the publisher (ask gemini or claude about objections to the ACM TAPS system for example).
Our document class runs some checks at compile time to make sure that authors don’t muck with the default style, and authors get a headsup while they are writing their article about things they should adhere to. That’s why I had asked about support for something like \@ifpackageloaded. We could probably get by without it, but I’d like to make a template for authors that flag things for them during their authoring process rather than having them upload their final document only to find out that they used unauthorized packages.
I haven’t looked closely at the structure of the HTML output from typst, but it’s easy for me to believe that it produces cleaner output than any of the LaTeX tools. The basic output produced by typst would have to be modified by modifying the standard html template though. The HTML that is produced for an article would typically be embedded in a larger page that has the navigation and styling of the journal, so that for example the title of the article would be an <h2> tag instead of an <h1> tag. There are a bunch of other things to check to make sure that the HTML is mobile-ready. It certainly doesn’t come for free from typst (it also doesn’t come for free with latexml which is used by arxiv.org). I spoke to a couple of guys on the PreTeXt team and they also haven’t addressed this template issue.
The ORCID part is part of their pipeline to authenticate authors when they upload their source code. Different authors use different templates to control how the ORCIDs are displayed in the HTML or PDF. It’s just one of the many features of their pipeline, but the ORCID organization has guidelines that stipulate how the ORCID should be displayed based on whether it was properly authenticated or not. Personally I’m not a big fan of ORCID, but some publishers are being tricked into accepting papers that were not really authored by the people that were claimed, and they are looking for an authentication technique to detect it.
I doubt that any of our authors would use a structure other than BibTeX in the short term. DBLP, AMS, and others already export millions of well-curated records in BibTeX and that’s what almost all of our authors use. BibTeX is a festering mess of a standard, but it’s still the defacto standard in mathematical disciplines.
I came back here to find out if there had been any progress on publishers accepting typst, and I now have a volunteer who wants to write a template for us. I may encourage him to try, but I think it would be a huge amount of work to integrate it into our pipeline because there is a lot more to it than he may realize.
1 Like
really glad that I inserted an H1 tag into my response. 
I’d fix it but I’m on an airplane with my laptop.
I like the absurdity of the formatting, but yeah, you might want to fix that.
As for the person volunteering, I’d say might as well let them try to figure out what is and isn’t going to be a pain in adoption.
I just want to make sure that the PhD student understands how complicated it is before he starts on it. I have other projects he could potentially work on that might have higher impact, but ultimately it’s his choice. If nothing else he would learn how to write better typst so that’s probably a good thing.
In retrospect the inserted h1 tag kind of illustrates the kind of nonsense that authors stick into their papers. I think I fixed it now.