data.html

<!DOCTYPE html>
<html lang="en-AU">
<head>
<meta charset="utf-8">
<meta name="viewport" content="width=device-width, initial-scale=1">
<title>Where the numbers come from · The Straddie Twins</title>
<meta name="description" content="Packs, feeds, receipts, the wanted list and the open questions: how each twin comes to know a thing, and what it does when it cannot.">
<meta name="color-scheme" content="dark">
<meta name="theme-color" content="#050912">
<link rel="preconnect" href="https://fonts.googleapis.com">
<link rel="preconnect" href="https://fonts.gstatic.com" crossorigin>
<link rel="stylesheet" href="https://fonts.googleapis.com/css2?family=Exo+2:wght@500;600;700;800&family=Manrope:wght@400;500;700&family=JetBrains+Mono:wght@400;600&display=swap">
<link rel="icon" href="data:image/svg+xml,%3Csvg xmlns=%27http://www.w3.org/2000/svg%27 viewBox=%270 0 32 32%27%3E%3Cdefs%3E%3ClinearGradient id=%27g%27 x1=%270%27 y1=%270%27 x2=%271%27 y2=%271%27%3E%3Cstop offset=%270%27 stop-color=%27%235ee7ff%27/%3E%3Cstop offset=%27.5%27 stop-color=%27%238f7bff%27/%3E%3Cstop offset=%271%27 stop-color=%27%23ffd166%27/%3E%3C/linearGradient%3E%3C/defs%3E%3Crect width=%2732%27 height=%2732%27 rx=%278%27 fill=%27url(%23g)%27/%3E%3Crect x=%277%27 y=%277%27 width=%2718%27 height=%2718%27 rx=%275%27 fill=%27%23050912%27/%3E%3C/svg%3E">
<link rel="stylesheet" href="assets/site.css">
</head>
<body data-page="data">
<header data-topbar></header>

<section class="page-hero">
  <div class="hero-photo" data-photo="detail-eighteen-mile.jpg" aria-hidden="true"></div>
  <div class="wrap">
    <p class="step-tag" data-step></p>
    <p class="eyebrow">Where the numbers come from</p>
    <h1>Most numbers name their source.</h1>
    <p class="lede">Both twins are built from what the island has published about itself.</p>
  </div>
  <p class="credit">Copernicus Sentinel-2, 13 July 2026 &middot; <a href="about.html#sources">sources</a></p>
</section>

<section class="section">
  <div class="wrap">
    <p class="eyebrow" data-reveal>The first build</p>
    <h2 data-reveal>Four ways a fact gets in.</h2>
    <p data-reveal>A pack is a data file of records, one subject per file.</p>
    <ol class="numbered" data-reveal>
      <li><b>Committed packs.</b> Twelve in the README's count, 1.9 megabytes, about three thick paperbacks of text. Around fourteen hundred records, 1,100 source fields, 760 source addresses. Most marked high, medium or low confidence. Nearly everything on screen comes from here.</li>
      <li><b>Documents with page citations.</b> A tool reads a named public document, pulls out the island lines, and writes candidates: page number and sentence, word for word. It never writes a pack. A person decides what is true; nothing unread becomes data. Two Redland City Council documents are registered, none promoted when the README was written.</li>
      <li><b>Contributions from islanders.</b> The running build has one. A seafood business at Amity Point still shows in several directories. A resident said it had closed. Recorded 6 May 2026, kept as closed, so a stale listing cannot put it back. No form, no portal, somebody typed it in.</li>
      <li><b>Synced feeds, with a date.</b> A person runs a sync offline. It reads a real source and writes a stamped feed into the repository. The running twin fetches nothing itself. Each synced record says when, from what source, and how fresh that source claimed to be. Four feeds: weather, wildlife rescue, the noticeboard, the events engine. The twin reads weather only; it goes stale after a day.</li>
    </ol>
  </div>
</section>

<section class="section">
  <div class="wrap">
    <p class="eyebrow" data-reveal>Trading hours and the cut-off</p>
    <h2 data-reveal>Three words for an opening time.</h2>
    <div class="split">
      <div>
        <p data-reveal>The business pack grades trading hours three ways.</p>
        <div class="cards">
          <div class="card" data-reveal><span class="chip real">Published</span><h3>The business published them itself.</h3><p>On its own site or page.</p></div>
          <div class="card" data-reveal><span class="chip modelled">Listed</span><h3>Someone else published them.</h3><p>A council, a tourism body or an island directory.</p></div>
          <div class="card" data-reveal><span class="chip proposed">Estimate</span><h3>Nobody published them.</h3><p>A pattern for the simulation, never shown as fact about a real street.</p></div>
        </div>
      </div>
      <div>
        <p data-reveal>A record below the confidence cut-off is not shown. The build calls that held: invisible, no note attached. Thirty-one businesses were marked unconfirmed at 10 August 2026, real things nobody had checked lately.</p>
        <p data-reveal>The build blurs locations on the way in, never at display time. Wildlife rescue coordinates round to about 110 metres, their times to the hour. Holiday lettings round to 250 metres. A display filter, the build says, is "one bug away from publishing the thing it was hiding". The intake refuses contact details and home addresses.</p>
      </div>
    </div>
  </div>
</section>

<section class="section">
  <div class="wrap">
    <p class="eyebrow" data-reveal>The wanted list</p>
    <h2 data-reveal>What the first build does not know.</h2>
    <p data-reveal>The twin lists its own gaps, counted from the packs. Trading hours nobody published. Premises with no coordinates. Businesses nobody could confirm still trade. One cafe has two published street addresses, so its coordinates stay empty until a local says which. Fill a gap and your name goes on the record.</p>
    <div class="table-wrap" data-reveal>
      <table>
        <thead><tr><th scope="col">Pack</th><th scope="col">What is in it</th></tr></thead>
        <tbody>
          <tr><td class="k"><b>geography</b></td><td>Ground, coast, roads. Missing roads fail some vehicle trips.</td></tr>
          <tr><td class="k"><b>places</b></td><td>Townships, beaches, headlands, jetties, halls.</td></tr>
          <tr><td class="k"><b>businesses</b></td><td>Every business, graded hours. Closed stays closed.</td></tr>
          <tr><td class="k"><b>ecology</b></td><td>Koalas, whales, shorebirds, marine life, vegetation, dunes, lakes.</td></tr>
          <tr><td class="k"><b>residents</b></td><td>The pattern generated residents come from. No real person.</td></tr>
          <tr><td class="k"><b>transport</b></td><td>Ferry timetables, roads, beach closures around high water.</td></tr>
          <tr><td class="k"><b>civic</b></td><td>Sixty-two levers, who decides each, council process, budget.</td></tr>
          <tr><td class="k"><b>events</b></td><td>What is on, organiser, source.</td></tr>
          <tr><td class="k"><b>narrative</b></td><td>Seeds for the story layer.</td></tr>
          <tr><td class="k"><b>soundscape</b></td><td>Generated sound: surf, wind, birds, engines.</td></tr>
          <tr><td class="k"><b>subterranean</b></td><td>Works below the sand, marked proposed everywhere.</td></tr>
          <tr><td class="k"><b>legislation</b></td><td>Which Acts bind which person.</td></tr>
          <tr><td class="k"><b>lore</b></td><td>Acknowledgement wording and the build's bans, read by its checker.</td></tr>
        </tbody>
      </table>
    </div>
    <p class="faint" data-reveal>Thirteen files sit in the data folder today; the README's twelve predates the lore pack. Beside them: a feeds folder and a contributions folder.</p>
  </div>
</section>

<section class="section">
  <div class="wrap">
    <p class="eyebrow" data-reveal>The second build</p>
    <h2 data-reveal>A receipt for every fetch.</h2>
    <div class="split">
      <div>
        <p data-reveal>The second build fetches only through a connector, a tool run from the terminal. It writes a receipt beside every raw file: the address, the time in island time, the server's answer. Then the size in bytes, the seconds taken, the licence as read. Plus a hash, a fingerprint of the exact bytes. A failed fetch writes nothing, and the tool says: do not type in a timetable instead.</p>
        <p data-reveal>Everything a connector fetches lands in a sealed inbox, out of the repository's history. A check tests that by asking the version-control system, not a setting. Five connectors exist: OpenStreetMap, TransLink, past weather, the 2021 census, the Atlas of Living Australia. Only the bus pack has been committed.</p>
        <p data-reveal>Every record points into a sources file by id, never free text. Each source gives title, publisher, the exact request. Then the date fetched, the licence text as read, the date read, the credit owed.</p>
      </div>
      <div>
        <div class="card gold" data-reveal>
          <p class="eyebrow gold">Recorded as missing</p>
          <h3>The tide table nobody obtained.</h3>
          <p>Nobody has obtained a published table of harmonic constituents for Dunwich, the standard port: the numbers that give you the tide at any hour. The tide pack has none, plus a record saying so: "not obtained", count zero, the source, the date. The ferry page prints "contested or unknown" with the date.</p>
        </div>
        <div class="card" data-reveal>
          <p class="eyebrow">The five packs</p>
          <ul class="ticks">
            <li><b>crossings:</b> the boat timetables, transcribed and sealed.</li>
            <li><b>transport-gtfs:</b> the island bus feed, with its receipt.</li>
            <li><b>tide:</b> zero constituents and the absence record.</li>
            <li><b>_sources:</b> every source by id, with its licence.</li>
            <li><b>lore:</b> the acknowledgement wording, byte for byte the first build's file.</li>
          </ul>
        </div>
      </div>
    </div>
  </div>
</section>

<section class="section">
  <div class="wrap">
    <p class="eyebrow" data-reveal>Open questions and an open decision</p>
    <h2 data-reveal>What the crossings pack cannot answer.</h2>
    <p data-reveal>The crossings pack stores its open questions as records: the question, what is known, the possibilities, confidence unknown. None answered yet.</p>
    <ul class="ticks" data-reveal>
      <li>Whether the vehicle barge is booked or turn up and queue.</li>
      <li>Which vehicle sailings SeaLink adds beyond the printed timetable, shown only on its booking page.</li>
      <li>Whether today is a public holiday. The build has no calendar of them, and four Gold Cats Flyer departures stop on one.</li>
      <li>What changes in the school holidays. No source in the build has that timetable.</li>
      <li>How long the walk-on crossing takes: neither passenger operator publishes it.</li>
      <li>Which mainland berth the Flyer uses, and how many berths Junner Street has.</li>
      <li>Why two of fourteen mainland departures differ between the two published sources.</li>
    </ul>
    <p data-reveal>OpenStreetMap data comes with the Open Database Licence's share-alike rule: use it and you share your own work back. The owner decides that, not the build. Until he does, the OpenStreetMap, weather, census and ecology packs stay in the inbox, blocked by a gate rule. A licence questions document sets out the options. The bus pack is Creative Commons Attribution, so it did not wait.</p>
    <div class="hero-actions" data-reveal>
      <a class="btn primary" href="second-build.html">Inside the second build <span class="arrow" aria-hidden="true">&rarr;</span></a>
      <a class="btn" href="ferry-page.html">The ferry page</a>
    </div>
  </div>
</section>

<footer class="site"><div class="wrap"><span class="sig">Luke Catalyst Nathan Hayes x Claude · Minjerribah, 2026</span><span><a href="about.html">About and licence</a></span></div></footer>
<nav data-rail aria-label="The walk"></nav>
<script src="assets/components.js"></script>
<script src="assets/site.js"></script>
</body>
</html>