Cool URIs can be ugly (2023)

Original Article ↗ Hacker News Discussion ↗

Cloudflare Pages’ automatic, permanent redirects from `.html` files to extension-less URLs have reignited debate over what makes web addresses “cool” and stable. Commenters weigh the benefits of clean, technology-agnostic URLs against the practical simplicity of exposing `.html`, worry about hosting providers mutating URL schemes without opt-out, and revisit the W3C ideal that links should remain valid for decades. The thread broadens into questions of link rot, redirects, REST-style paths vs query parameters, and whether exposing implementation details like file extensions meaningfully harms long-term maintainability.

Cloudflare / GitHub Behavior and Redirects

  • Cloudflare Pages automatically redirects .html to extension‑less URLs and treats them as permanent; some see this as a critical, invasive “feature” that should be opt‑in.
  • Others argue this is long‑standing behavior in many static hosts and web servers, driven by the mismatch between filesystem extensions and URL content types.
  • Criticism centers on the permanence (301s) and lack of configuration options; supporters frame it as a product decision, not a bug.
  • GitHub Pages has similar mapping rules (/foo → foo.html, folder → folder/ → folder/index.html), but generally without permanent redirects.

Cool URIs, Extensions, and Future‑Proofing

  • The W3C “Cool URIs don’t change” stance against .html is revisited.
  • Some agree extensions expose backend details and risk future breakage; others say .html is stable enough and unlikely to change for typical sites.
  • A common pragmatic view: if you choose .html and keep it stable, that becomes the “cool” URL; switching backends should adapt to existing URLs, not vice versa.
  • Some frameworks even tie extensions directly to MIME types (.json, .xml, etc.) as a debugging aid.

What Counts as an “Ugly” URL

  • Many participants reserve “ugly” for long, parameter‑laden or hash‑encoded URLs, not simple /year/slug.html.
  • Historical examples include comma‑infested CMS URLs, JSON blobs in fragments, tracking query strings, and complex enterprise app URLs.
  • Static paths with IDs plus slugs (/category/123/post-name) are seen as a good compromise: stable identifier plus human readability.

Redirects, Link Rot, and Responsibility

  • One camp: site operators have no obligation to preserve URLs forever; “cool URIs” rhetoric can feel moralizing.
  • Another camp: maintaining redirects is low effort and courteous; link rot is primarily caused by targets changing, not link authors.
  • Some suggest even static‑site workflows can auto‑generate simple redirect pages or meta‑refreshes.

Implementation Details and Monitoring Concerns

  • Popular pattern: /blog/slug/ directory with index.html, sometimes used per‑post to hold related assets.
  • Content negotiation and alternative formats (Markdown, multilingual variants) often clash with static generators’ assumptions.
  • Observability practitioners dislike putting user data in path segments, since it complicates metrics and logging; they prefer query parameters.

URIs vs URLs

  • There’s a short historical recap of URIs, URLs, URNs, and their RFC lineage.
  • Many agree that in everyday web work “URL” is the only practically useful term, even if “URI” is technically more general.