The promise would be that in HTML, everything is reformatted to look great on any line length.

But there is a kind of integration attack going on: Once we have the HTML version, why not enhance the paper with things PDF can't do? After all, HTML5 offers limitless possibilities to make the 3D figures interactive, run simulations, have the examples to be toys you can play with, etc.

Once these seemingly harmless new features are there, who would want to go back to the PDF version, which is a
2/4

static, cumbersome and arcane? Once we are there, we will definitely need versioning of publications because if papers become software, they will have bugs and need bug fixes.

I feel very insecure about this future. I am unhappy with how static and old-fashioned it is of us, to stick to this A4-PDF-paper format, for documents which are consumed on laptop and iPhone screens, and it never really fits and scrolling and zooming and whatnot.
3/4

On the other hand, I want to be done with a paper at some point. If publications turn into software, they will just need a lot of maintenance, forever. Maybe it can be solved with open-source culture, maybe future generations will take care of the bug fixing for my papers, but why would they and how would the get credit for this work in an academic system that only rewards new things and not the maintenance?

So what do you think about HTML arXiv papers?

4/4

@tomkalei @buerviper personally I never author PDFs. Too much like dead paper. Presentations on my website are pure HTML and ready for full interactivity using my own open source library:

https://analyticphysics.com
https://paulmasson.github.io/mathcell/docs/

Of course I'm not part of the academic system, so I can implement whatever I feel appropriate. The MathCell library replicates Mathematica's Manipulate command in pure JavaScript, so it runs just by loading the page. Wouldn't have it any other way.

When I make major changes to a page I also note that at the bottom. More detailed tracking of changes is easily implemented with something like GitHub.

@buerviper @paulmasson If research results are versioned like this, would there not be incentives to publish 0. Versions way too early, fix things as you go, and import things that are bad in software development into science and engineering?

Also how would citing work if publications are not static but are updated?

Also some parts of math are super conservative. State of the art results are 50+ years old. Would that kind of quality still be possible in a much accelerated publishing culture?