返回资讯中心
外部精选
软件工程
#开发者实时流

Aligning a Personal Site with Indieweb and Web Standards

TITLE: ALIGNING A PERSONAL SITE WITH INDIEWEB AND WEB STANDARDS

Hacker News 最新jruohonen2026/10/18 · 2026/10/1819 分钟阅读

以下正文同步自 Hacker News 最新,版权归原站所有,已转换为易读排版。

TITLE: ALIGNING A PERSONAL SITE WITH INDIEWEB AND WEB STANDARDS DATE: 30 SEPTEMBER 2026 TAGS: INDIEWEB, MICROFORMATS, WEBMENTION

Krzysztof Lewczuk

RECLAIMING THE DOMAIN

I recently decided it was finally time to kill the "coming soon"-like placeholder on my main domain, htsh.pl. The initial plan was simple: throw up a minimalist personal/semi-business vCard and call it a day.

But scope creep set in. That simple vCard evolved into a full-fledged hub featuring a blog, project showcases, and a repository for technical tidbits and research notes - all wrapped in a dark-mode, teletext-inspired aesthetic.

Building a site from scratch in 2026, it's easy to assume you already know how the internet works: buy a domain, provision hosting, write some markup. But building something meant to last and interoperate with the broader web means revisiting the actual standards. If you're starting this journey, the IndieWeb Getting Started guide is the perfect launchpad.

Here's what I ended up auditing and adding to htsh.pl - from the boring web basics to the IndieWeb-specific pieces.

Before diving into code, it helps to know who's driving the underlying rules of the web. A few key organizations do most of the heavy lifting:

Before adding any IndieWeb-specific markup, I went through the less exciting parts of the site first. Most of them don't need a framework or another service - they just need to be done properly.

My policy is intentionally permissive. I don't mind the site being indexed or used by AI systems, so I explicitly allow the relevant agents and tokens instead of leaving that policy implicit behind User-agent: *.

None of this is particularly interesting technology. That's probably the point: the boring baseline should already be in place before adding another layer of metadata and protocols on top.

This is where standard web design turns into being a good citizen of the open web. Instead of relying solely on platform algorithms, IndieWeb is about owning your data on your own domain first.

To get there, htsh.pl runs a dual-strategy: microformats2 for IndieWeb tooling, and Schema.org JSON-LD for search engines. They overlap in purpose but serve different readers, and both coexist fine on the same page.

Microformats are just HTML class names - h-card, h-entry, p-name, u-url - that make existing markup machine-readable, no API or schema registry required. On htsh.pl:

The h-card describes who the page belongs to, but it doesn't by itself connect that identity to profiles elsewhere. That's where rel="me" comes in: it says that another URL represents the same person.

On htsh.pl, the GitHub and LinkedIn links carry it directly:

<a href="https://github.com/krzlew" rel="me">GitHub</a> <a href="https://www.linkedin.com/in/krzysztof-lewczuk/" rel="me">LinkedIn</a>

The useful part is the backlink. My GitHub profile's Website field points back to htsh.pl, creating a reciprocal relationship between the two URLs - that's what makes rel="me" checkable by other tools, not just declarative.

Microformats2 covers the IndieWeb side of the site. For search-engine structured data, every post also includes a Schema.org application/ld+json script block, which is what Google recommends for most structured-data implementations, so crawlers know exactly who wrote the article and when it was published. Both markup styles read the same underlying facts - they just target different audiences, and neither one replaces the other.

I already follow the basic POSSE idea - Publish (on your) Own Site, Syndicate Elsewhere - although the syndication part is manual.

Posts are published on htsh.pl first, and some of them are then adapted and posted to LinkedIn. There is no automation or publishing API involved; I just treat my own site as the canonical copy and LinkedIn as the distribution channel.

What I haven't added yet is u-syndication. For posts that have a corresponding LinkedIn copy, I could point it at that post's permalink and make the relationship explicit in the markup. For everything else there is nothing to advertise, so adding empty or invented syndication links would just be checklist theater.

Separately, the site publishes Open Graph (og:image, og:title) and Twitter Card metadata using the pixel avatar. That's unrelated to POSSE itself - it controls how links render when someone shares them in places such as Discord, LinkedIn or Mastodon.

To support following the site without a platform in the loop at all, it also serves standard RSS and Atom feeds (/feed.xml and /atom.xml). All roads lead back to the domain.

Before calling this done, test it against the checklists the IndieWeb community built for exactly this purpose:

I ended up with two separate interaction paths:

Webmention is a decentralized link-notification protocol that can carry replies, mentions and other cross-site interactions. Two lines in <head> point at a receiving endpoint:

<link rel="webmention" href="https://webmention.io/htsh.pl/webmention"> <link rel="pingback" href="https://webmention.io/htsh.pl/xmlrpc">

The endpoint itself is hosted by webmention.io - free, and it doesn't do anything until the domain is claimed there via IndieAuth. Claiming works via a RelMeAuth-style check: IndieAuth confirms a domain and an external profile belong to the same person by looking for exactly the reciprocal rel="me" backlink set up earlier, on GitHub in my case. Webmention itself has no identity requirement at all, though - at the protocol level, a source just tells a target that it links there, and the target fetches the source URL to confirm the link actually exists. rel="me" only matters for this claiming step, not for the act of sending or receiving a Webmention. Once claimed, when someone links to a post here from a Webmention-aware site, their server notifies the webmention.io endpoint advertised by my page, and the mention shows up in a feed I can check.

I deliberately skipped webmention.io's webhook option, which pushes new mentions to a callback URL instead of me polling the feed - that needs a server to receive the POST, and this is a static Eleventy site with no backend. Standing up a serverless function just to avoid checking an Atom feed occasionally isn't worth the infrastructure for the mention volume a small personal site actually gets.

Giscus covers traditional, human-readable comments right on the page, backed by GitHub Discussions. Webmention and Giscus aren't competing with each other here; Giscus is for readers who want to leave a comment directly, Webmention is for readers who wrote their own response elsewhere and want this site to know about it.

Once the site is live, you need to tell the web it exists:

The useful cutoff turned out to be fairly simple: anything that improves interoperability for almost no runtime cost stays. Semantic markup, microformats, feeds, rel="me", structured data and Webmention discovery all fall into that category.

The line moves once another service, credential, deployment target or backend is involved. Syndication happens manually to LinkedIn today, so there is no u-syndication markup yet - adding it once that copy's permalink actually exists is cheap, but automating the syndication itself is not, so it stays manual. Webmention volume doesn't justify a webhook receiver, so I poll the feed. There is no point adding infrastructure just to tick another IndieWeb box.

That is probably the part of this rebuild I want to keep: use the standards where they solve an actual problem, and stop when implementing the next one becomes checklist theater.

正文由 探How 从来源站点 RSS 同步,内容未经改写;遇到排版缺失或需要图片、视频时请以原文为准。

打开原文
去社区讨论 →