Skip to content

How to build this site

Wanna steal this website? You’re allowed to! Just remember that it’s licensed under the CC-BY-SA (for content) and GNU Affero GPL Version 3 (for code) licenses.

Once you put a few binaries in your $PATH, building is just a make invocation away.

Repository mirroring

I push the source code to the following forges:

Clone the repository from any of those locations.


Build-time dependencies

  • Hugo. I usually use the most recent version of Hugo at the time of publishing, but it should work with any version of Hugo v0.116.0 or later (v0.116.0 had an improvement to where that I might use). I build Hugo with the nodeploy build-tag for a smaller binary and faster build-times.

  • A make implementation. This site works with bmake (from NetBSD and FreeBSD) and GNU Make 4 or later. OpenBSD make (“omake”) should work too, but I haven’t tested it.

  • Git. Hugo uses Git for generating modification timestamps.

  • curl, for fetching some webring code and all my webmentions. Fetching Webmentions requires authentication. When running locally, it invokes pash, my password manager; when running in CI, it reads a file for a secret. You may have to modify scripts/ to avoid this.

  • POSIX utilities: grep, find, sed, POSIX-compliant /bin/sh, etc. Tested to work with Busybox and GNU Coreutils. Shell scripts run with Zsh, Busybox sh, and the Debian Almquist shell.

Before deploying, I use some tools for post-processing:

I also apply static compression at high levels, using the following tools:

  • Efficient Compression Tool (ect). It’s like zopfli but more efficient and faster. If you don’t have it installed, it should be trivial to edit scripts/ to replace ect with gzip or zopfli.

  • Brotli.

I package xmllint, hugo, brotli, and ect as statically-linked binaries in a tarball. You can download this tarball from You’ll need to install make, git, POSIX utilities, and curl yourself. These should be fairly ubiquitous; there’s a good chance that you already have them.

Other dependencies

To deploy, I use rsync with SSH and zstd support.

Other lints/checks I run also use additional command-line utilities such as sd, htmlq, and a version of xargs that supports the -P flag (nearly all versions of xargs do, but it’s not POSIX). I run most npm packages using pnpm -s dlx (similar to npx).

Quick linters

  • Stylelint

  • html-validate

  • A very recent build of the W3C’s Nu HTML checker to validate the HTML and XHTML, available on your $PATH as vnu. I have a very simple shell-script wrapper for this that invokes java -jar. You may have to update the CSS-Validator submodule to its latest commit to avoid false positives.

  • jaq, to filter false-positives from the Nu HTML checker and to verify that JSON files parse. You can replace jaq with jq without any issues.

  • HTMLProofer, version 5 or later. Requires Ruby.

  • W3C feed validator (requires Python)

Slow linters

These linters use headless browsers and take several minutes to validate my whole site.

  • Axe-Core, using the CLI and a headless browser (Firefox or Chromium).

  • IBM Equal Access Checker, using the CLI and a headless Chromium. Runs on a patched version of the site with all instances of content-visibility removed from the stylesheet. I do this to work around a false-positive, reported upstream.

Testing the deployed site

  • WebHint (normally runs on every page in my sitemap with a headless browser. Abandonware but still works.)
  • Lighthouse
  • redbot

Build instructions

  • To just build the HTML: make hugo

  • To build the polygot formatted HTML and XHTML: make xhtmlize

  • To lint and validate: make xhtmlize lint-local

  • To build everything and compress: make xhtmlize compress copy-to-xhtml

  • To deploy the clearnet site and corresponding Tor hidden service: make deploy-prod deploy-onion.

lint-local and deployment tasks support limited parallelization with -j.