Deprecate Middleman

[ikuwow]

Middleman ecosystems are somehow inactive now.

[ikuwow]

Research summary as of 2026-07-06. Maintenance status verified via GitHub API; migration candidates verified via official docs and GitHub. Decision for now: stay on Middleman.

Middleman ecosystem maintenance status

• middleman/middleman: not archived. 64 commits in the last year, but almost all dependabot. Human commits: Ruby 4.0 support (2026-02), v4.6.3 prep (2026-02), HAML 7 fix. v5 has been stalled at rc1 since 2019.
• middleman-blog: alive. Ruby 4 CI support (2026-02), v4.2.1 tagged 2025-11.
• middleman-syntax: alive. Rouge 4.x / Ruby 4 support (2026-02).
• middleman-livereload: slow but alive (last commit 2025-07).
• middleman-minify-html: dead (no commits since 2020-08). Removal tracked in Remove unmaintained plugins: middleman-minify-html and middleman-favicon-maker #926.
• middleman-favicon-maker: dead (last commit on master 2017-11). Removal tracked in Remove unmaintained plugins: middleman-minify-html and middleman-favicon-maker #926.

Assessment: the core and blog stack are in life-support maintenance mode (Ruby compatibility updates only, no feature work). Not an imminent risk. The real risk is a future Ruby or dependency incompatibility going unfixed.

Migration candidates (if/when triggered)

Candidate	Maintenance	Migration cost	Notes
Astro	Very active	Medium	astro-embed may replace the oEmbed extension; bare-URL auto-embed on plain .md unverified (may require MDX)
Eleventy	Active	Medium	permalink front matter reproduces /entry/{title}/ directly; eleventy-plugin-embed-everything is maintained
Bridgetown	Monthly releases, small community	Low	ERB is the default template engine; layout assets and Ruby knowledge carry over; bus-factor risk
Hugo	Very active	Medium-high	Go templates are a different paradigm; oEmbed auto-conversion needs custom render hooks
Zola	Moderate	Medium-high	Taxonomy URL format is fixed; Japanese tag URL reproduction unverified
Jekyll	Stagnant (no release in 1.5 years)	-	Rejected: contradicts the goal of moving to something actively maintained

Hard constraints for any migration

• URL preservation is mandatory: utterances comments are keyed by pathname (issue-term="pathname" in source/layouts/article_layout.erb), so changing URLs orphans all existing comment threads
• Live URL formats: /entry/{slug}/, /tags/{tag}/ (raw Japanese path segments), /archive/{year}/, /feed.xml
• Slugs come from the post directory name ({year}-{month}-{day}-{title} -> {title} part), not from the front matter title (which is Japanese); a migration script must write explicit slugs/permalinks per post
• Front matter needs normalization: date: YYYY-MM-DD HH:MM JST (non-ISO) and comma-separated tags strings (28 posts have multiple tags)
• 5 posts have published: false and must map to the target's draft mechanism
• The custom oEmbed extension covers Twitter/X only; baking the cached embed HTML into posts during migration would remove the oEmbed dependency entirely and neutralize that criterion
• S3 + CloudFront deploy and CircleCI structure are SSG-agnostic; only the build command changes

Decision and triggers

Stay on Middleman for now. Suggested migration triggers:

• middleman or middleman-blog breaks on a future Ruby version and stays unfixed for months
• a required gem becomes incompatible with no workaround
• writing/tooling friction becomes a real cost

First candidates to PoC when triggered: Astro and Eleventy (acceptance criteria: exact reproduction of /entry/{slug}/, Japanese tag URLs, /archive/{year}/, /feed.xml, zero diff against the production sitemap URL list).