I don't always use LaTeX, but when I do, I compile to HTML (2013)

Original Article ↗ Hacker News Discussion ↗

Efforts to convert LaTeX documents into high-quality, mobile-friendly HTML are maturing but still face gaps in layout fidelity, math support, and package coverage. Commenters compare tools like LaTeXML, Sphinx, Pandoc, Typst, and Quarto, weighing the trade-offs between PDFs’ fixed, precise typography and HTML’s reflowable, accessible format. Many conclude that LaTeX remains hard to replace for complex, long-form scientific work, even as hybrid workflows and newer markup systems gain traction for web publishing and easier authoring.

LaTeX → HTML conversion

  • Original post revisited LaTeXML; users share command-line recipes (latexml, latexmlpost, latexmlc) and note it still has spacing/line-break issues but is improving.
  • arXiv has adopted LaTeXML for in-house HTML, with many reported bugs being actively handled.
  • Another approach (lwarp) redefines LaTeX macros to emit HTML, using the TeX engine as the only true parser; it’s used successfully on very complex docs (e.g., TikZ manual).
  • PDF→HTML tools like pdf2htmlEX impress visually but are criticized as “HTML that looks like a PDF,” not mobile-friendly, reflowable content.

Alternative authoring stacks

  • Sphinx + reStructuredText praised as a powerful, underrated system: multi-format output (HTML, LaTeX→PDF, epub), math support, Graphviz/Mermaid, custom directives, requirements management.
  • People describe pipelines: Markdown → pandoc → LaTeX/PDF/HTML; RMarkdown/Quarto → HTML/PDF/docx; Jupyter/Quarto for notebooks and websites.
  • Extensions exist for citations (Zotero + Better BibTeX + sphinxcontrib-bibtex), code execution, and auto-included tables/CSV.

Typst and other formats

  • Typst gets enthusiasm as a modern LaTeX-like typesetting system with fast live recompilation and pleasant syntax, but lack of HTML export is repeatedly flagged as a blocker for this specific use case.
  • Some see “X but in Rust” rewrites as a trope, yet concede many Rust-based tools (including Typst) are meaningful improvements.
  • AsciiDoc and Asciidoctor are discussed as cleaner than reST by some, too complex or with odd table syntax by others; ecosystem maturity seen as still in flux.

PDF vs HTML, and device considerations

  • Strong camp: LaTeX+PDF is ideal for fixed, typographically precise scientific docs, book manuscripts, and long-term archiving; readers are expected to use large screens or printouts.
  • Opposing view: PDFs (especially multi-column) are painful on phones; HTML is better for reflowable reading, though achieving LaTeX-level typography and structure in HTML is hard.
  • Some argue that trying to fix layout across all devices is misguided; others counter that TeX’s page-layout strengths are exactly the point.

Structure, templates, and academic workflow

  • LaTeX remains dominant because publishers provide exact templates (APA, IEEE, journal-specific styles), and content can be dropped into these without reinventing formatting.
  • Users of Sphinx/Markdown want similarly robust, maintained templates for academic standards, not just “roll your own” theming.
  • Automatic indexing, cross-references, and large-document management are cited as areas where LaTeX still “just works.”