[tex4ht] why arxiv.org choose LaTeXML and not tex4ht for making HTML pages?

Deyan Ginev deyan.ginev at gmail.com
Tue Mar 8 22:09:27 CET 2022

Hi Bill,

The idea of a profiled, or constrained, use of LaTeX has revealed
itself to be self-defeating during the last decade in a somewhat
amusing way.

The moment one decides on a fixed set of declarative markup
constructs, as to establish a "profile", there are instantly better
tools to use in 2022 than the TeX engine itself.

The standard example: If we are restricted to sectioning, lists and
basic images and tables, then basic markdown is smaller, simpler to
write, and easier to map into any target format.
Which also makes it straightforward to be supported by developers for
all kinds of modern applications.
As the document complexity grows, larger markdown dialects keep
appearing - Github-Flavored Markdown, Scholarly Markdown, Madoko,
Markua ... I'm sure there will be more.
The XML publishing world has its own small ecosystem of such experiments.

But the moment the first \def sneaks in, and we become
Turing-complete, a pandora's box opens that can not be closed
As long as the expectation is to have a TeX engine "proper", one will
always be able to escape an artificially imposed profile.
Advanced users of LaTeX will always have that expectation - and dodge
discipline - especially judging by recent arXiv submissions.

In the same breath, I also have to state:

The fully open, highest-difficulty, LaTeX problem needs to be tackled
today. At least on a software development level, if not on a Computer
Science one.
You may be familiar with the statistic that arXiv submissions continue
to increase:

As long as the general trend continues, projects like tex4ht and
latexml will have work left to do, as well as communities that need
Documents written in LaTeX today are valuable and need a path beyond PDF.


On Tue, Mar 8, 2022 at 2:44 PM William F Hammond <gellmu at gmail.com> wrote:
> On Sun, Feb 27, 2022 at 2:16 PM Michal Hoftich replied to Nasser Abbasi:
> [snip]
> > Are their technical reasons for this choice? What can LaTeXML do
> > which tex4ht can not?
> >
> I think LaTeXML uses a different approach than TeX4ht for the
> conversion, so while we can easily support custom commands and
> packages, or translate chunks of the document to pictures, we are also
> more likely to get fatal errors because of package clashes. So just
> using a package  in your document (for example, recently I had to fix
> the PDFX package), can lead to a fatal error. This can be difficult to
> debug for a lot of users, and even more problematic if you have
> thousands of documents from various users, all of them using random
> packages. You will just get a big number of documents that fail.
> Thousands of documents from various users using random packages with no umbrella of discipline is the reason for my suggestion that the LaTeX community should begin to use LaTeX profiles as outlined in my talk in San Francisco at TUG 2010.
> The author of a document using a LaTeX profile that is supported by both LaTeXML and TeX4ht would be able to have confidence that their [sic] document would pass correctly through both systems as well as through the system of any publisher supporting the profile.
>             -- Bill
> William F Hammond
> Email: gellmu at gmail.com
> https://www.facebook.com/william.f.hammond
> http://www.albany.edu/~hammond/
> 𝑺𝒖𝒑𝒑𝒐𝒓𝒕𝒊𝒏𝒈 𝒂𝒇𝒇𝒊𝒓𝒎𝒂𝒕𝒊𝒗𝒆 𝒂𝒄𝒕𝒊𝒐𝒏 𝒇𝒐𝒓 𝒕𝒉𝒆 𝒇𝒊𝒏𝒊𝒕𝒆 𝒑𝒍𝒂𝒄𝒆𝒔

More information about the tex4ht mailing list.