I run an event directory

You collect events from many places and list them in one spot: a city's listings, a fandom's convention calendar, a community's meetups.

What you're dealing with now

• You scrape or you type. Every source has its own format, or none, and each scraper breaks on its own schedule.
• Duplicates. The same event arrives from the organizer, the ticket shop and another directory, and you merge them by guessing from names and dates.
• Stale listings. A cancelled event often just disappears from its source, so you keep listing it.
• Your list stays on your site. People who want your events in their calendar copy them over one by one.

Collect from everyone, publish to everyone

• Organizers and ticketing sites publish documents, indexes and manifests you can collect from (I run a convention's website, I build a ticketing site). One crawler covers all of them.
• Duplicates come labelled. Every event has a stable id, and the same event on two sites carries an alternate relation, so you merge on data instead of guesses.
• Cancelled events say so. The spec asks publishers to keep them up with a status instead of deleting them.
• What you collect, you publish again. Calendar apps (I'm building a calendar app) can follow your directory, so your users get your picks in the app they already use. Other directories can build on yours the same way.

What's in it for your listings

• Dates as precise as the organizer knows them, so "June 2027" can be listed before the day is set, and sorted with the rest.
• Names in many languages, so one listing works for readers of each.
• Series and parts. An edition links to its series, and a meetup to the convention it's part of.
• Where it came from. Listings you write yourself point at their source.

What it costs you

You already crawl. This gives the crawler one format to read and a JSON Schema to check it with. Publishing again is the same document format, plus an index and a manifest.

How

Finding events

• Manifests at /.well-known/open-evnt/manifest on the sites you follow. Skip item types you don't understand, and respect auth. See the Manifest.
• Indexes list event URLs. Follow next, and dedupe items by href, since an item can show up on two pages. Use updatedAt or etag to skip events that haven't changed. See the Index.
• Web pages of single events can point to their document with <link rel="alternate"> (Link Discovery).

Reject documents with a v you don't support, or with fields the spec doesn't define outside of components (§8.2). The JSON Schema checks most of this, but not everything.

One event, many documents

Key everything by id (§2.2).

• Verify https ids. Fetch the id, following redirects. If you get back a document with the same id, it's verified. If not, treat the id as a claim from whoever published it.
• Collisions. Two documents with the same id that clearly describe different events aren't the same event. Don't merge them.
• Duplicates. The same event listed on two sites has two ids. An alternate relation says so. A relation is a claim by the publisher that wrote it. For more certainty, check that the other document links back.
• Series and parts. parent and child relations connect an edition to its series, or a meetup to the convention it's part of. The same back-link check works here: child for a parent, parent for a child.

Publishing again

• Mirroring a document as it is: keep its id. An id doesn't change when a document is copied to another publisher.
• Writing your own from an announcement, a poster or a web page: it gets your own id, a source component pointing at where you got it, and an alternate relation to the organizer's document if there is one.
• Keep components you don't understand when you re-publish (§6.1).
• List what you publish in an index and a manifest, the way a ticketing site does.

Listing

• Keep cancelled and postponed events listed, but don't show their dates as upcoming (§2.1).
• Sort by when periods start, with finer precision first on a tie (PartialDate §3.2).

Showing dates, languages and places works the same as in a calendar app.