I’m not sure if this should go in another category or not, but has the Typst team had a discussion with the arXiv maintainers about what would be needed to get experimental support for Typst submissions to arXiv?
I think the arXiv team is reasonably fairly conservative with changes, but it would be pretty amazing from a workflow perspective and I think really unlock a lot of academic use of Typst in general.
You can already upload Typst-generated PDFs on arXiv/HAL, right? However I guess that these preprint platforms need Latex sources if you want advanced features such as author/bibliography automatic scanning or watermark integration in the margins or first page. As far as I’m concerned, I don’t have a strong need for these features at the moment.
This is really not desirable in any way. There is substantial benefit in the arXiv not only being the center of scientific discourse but also being a public archive for the generating source materials.
As far as I’m concerned, I don’t have a strong need for these features at the moment.
Good for you, but that’s not relevant to this topic.
Sorry, but I struggle to understand the particular need you want to fulfill regarding arXiv. Apart from having the arXiv watermark on the PDF, what would be the other advantages of being able to send Typst sources to arXiv? I’m genuinely curious.
arXiv’s FAQ explains why they want the TeX source over PDFs: Requiring TeX when possible - arXiv info
While the article is about TeX and not typst, I’d imagine the reasons for preferring typst sources over PDFs (if/when typst is supported) would be would be the same.
As far as I know, there has been no official attempt to reach out to them on side of the Typst team. There was someone on Reddit who has tried reaching out a year ago, though.
Maybe its agood time to do so, especially with arXiv’s new HTML initiative, it’s a great way to collab into this feature of Typst and slowly start gaining traction from the publication community.
One of the arXiv maintainers was at the Typst Meetup last weekend, giving a talk on the topic. We and other attendees also had some productive discussions with him.
The full talk will be uploaded to our YouTube channel in the coming week or so, so stay tuned for that!
The video with that maintainer is now available to watch online:
What I gathered from his reasoning is mainly that they would have to reliably compile all submissions at any time in the future.
I would agree with him that the Typst Universe repository is changing too much at the moment to be able to have all compilation done entirely on their end, for every Typst submission ever. Especially considering one can have tens of revisions for a single plugin-based package, where each revision is in the megabytes.
But I would also argue that they could prohibit the use of Typst Universe packages altogether for the foreseeable future? I’m afraid that would eventually lead to people stealing packages, then submitting as part of their own code.
This was briefly discussed in the off-topic channel. This is basically a non-issue. CSL files can also be grabbed from GitHub - citation-style-language/styles: Official repository for Citation Style Language (CSL) citation styles. · GitHub.
The problem is fonts, mainly, I think. This exact issue can be clearly seen in this project of the recent Typst paper. The web app doesn’t have all the “usual” FOSS fonts, and some of them are not even possible to use due to size, i.e., Noto Serif CJK SC is a 23 MiB .otf font, while the limit is 20 MiB. This affects VCS-tracked files as well.
Another problem might be the complex pipeline that a work can have. Using the same paper example, the compilation requires 3 additional compilation for native PDF figures. So, if this is disallowed, I’d have to submit 3 PDF files as sources, instead of 3 Typst files as sources (for the PDF figures).
Well, and the fact that HTML is in the experimental state. Though their HTML accessibility program is also experimental/recent.
I looked at this today as I wanted to send on arXiv my first typst-authored papers (I have a dozen or so LaTeX papers there). Here is my current understanding of the topic:
arxiv does accept PDF-only submissions (to work with people using Word for authoring), which I was previously not aware of! My understanding is that moderators handle these manually and first check why no source is available, so it is not guaranteed that they are processed as promptly as usual submissions. I sent my Typst submission as a PDF-only submission this morning and will let you know how it goes. I think that it is important for the Typst community to use this approach (when they have actual scientific preprints, of course) and make it clear that it is a Typst-authored papers, so that arXiv gets a clear signal that there is a community demand for a more integrated process.
the reply Has Typst had a discussion with the arXiv maintainers? - #9 by hpcfzl above is the best public answer to this question (of contact between Typst and arXiv people). This is a talk from a senior software engineer at arXiv (Norbert Preining, also a former computer science academic and a texlive maintainer), who say that they are well aware of Typst and several people on the arXiv team are typst&rust enthusiasts. This is very promising (infinitely more than the previous reddit discussion), as it shows that the two groups are aware of each other.
For me as a researcher it would be important to be able to publish on arXiv, which is an important way to publicize one’s work in my community, and is also used as a storage/archival backend by some open-access journals (“arXiv overlays”): it is likely that at some point the technical possibility of supporting Typst papers on arXiv is going to influence the possibility for me of using Typst professionally. For the time being I am going to explore how much of this question is solved by the PDF-only pipeline (I suspect: not all of it, because the extra friction is going to cause issues), but I think it would be important to discuss better integration.
The questions that the arXiv team are asking about the long-term archival of Typst documents, not just their rendered output, and continued ability to re-render them, are important I think not just for arXiv. (I am moderately surprised that there is not a somewhat-simple question based on Docker images with a fixed operating system and a fixed Typst release, plus possibly a script to automatically download the package dependencies of a document and build a local, minimalist package-repository of exactly those packages.)
This was already done by a lot of people, relatively speaking. In the paper (p. 9, Section VII), we analyzed CS category and found around 100 PDF files that were produced by Typst (2023–2025). arXiv already knows about Typst (and you also say it later), but there are several issues that need to be addressed to get things going.