/* ==========================================================================
   reveal.css — the four build write-ups arrive as they are read.

   WHY THIS IS NOT `.reveal`. style.css already carries a reveal system, and
   index.html is the one page that uses it: `.js .reveal` sets opacity 0 and a
   20px offset as a RESTING state, and a script adds .is-visible when the
   element scrolls in. It produces the better effect and it costs something
   real. The five browser checks do not scroll — every one of them loads a
   page, settles it and measures what is there — and resting.mjs skips any
   element at opacity 0. Measured at the 1440x1000 the checks use: 28 of
   index.html's 34 reveal blocks are still at opacity 0 when it is measured,
   and 111 text nodes on the homepage are never contrast-checked at all.

   Extending that pattern to four more pages would have multiplied a defect
   the tripwire counts in CLAUDE.md exist to catch, and would have done it in
   the direction that hides rather than announces: the numbers would simply
   come back smaller.

   So these pages use the contract the cockpit's own arrivals use instead.
   The keyframe has a `from` and no `to`, and nothing is styled until the
   script says the element is on screen — so an element that is never
   triggered, on a page whose JavaScript did not run, or under a reader who
   asked for no motion, sits at its finished state rather than waiting to be
   animated into one. Nothing on these pages is ever measurable at opacity 0,
   and the counts do not move by a single measurement.

   WHAT IT COSTS, SAID PLAINLY. A resting state that is already finished can
   be SEEN before the arrival starts, and then the arrival blinks it away and
   fades it back. That is worse than no animation, and it was real: with the
   observer firing 8% of a viewport late, the cockpit's card painted fully
   opaque on screen for thirty-six frames, up to 70px of it, before its own
   animation caught up. The answer is the sign of the margin. reveal.js fires
   while the block is still below the fold, so there is no frame in which it
   is both visible and untouched. Scrolled slowly enough the arrival finishes
   before the block appears and the reader sees a settled page, which is what
   this site does below the fold today and is the right way for this to fail.

   TWO ATTRIBUTES, NOT ONE, AND THE PROTOTYPE PAGE IS WHY. The cockpit sets
   data-shown on its own surfaces, and prototype/dispatch-cockpit.html loads
   both files. A rule on [data-shown] alone would reach inside the cockpit and
   put a 20px page rise on a factor list that has its own bars to draw.
   ========================================================================== */

/* IT TRAVELS AND IT DOES NOT FADE, WHICH IS THE ONE PLACE THIS PARTS COMPANY
   WITH THE SITE'S REVEAL VOCABULARY, AND IT IS NOT A TASTE DECISION.
 *
 * The first version of this file did both, the site's own 20px rise under an
 * opacity, and it cost exactly one measurement on the prototype page. The
 * mechanism is worth writing down because it will outlive this file and it
 * catches anything observer-driven:
 *
 *   reachable.mjs presses a control, the press scrolls the page, and the
 *   scroll brings a block into the observation area. settle() waits on
 *   document.getAnimations() -- but an IntersectionObserver callback has not
 *   run yet, so there IS no animation to wait on, and settle returns. The
 *   callback fires afterwards, the arrival starts, and the check measures the
 *   first frame. In "the close call, compared" two blocks sat at opacity 0
 *   with data-shown already set: triggered, and one frame old.
 *
 * No amount of waiting fixes that from inside the page, and starting the fade
 * at 0.01 instead of 0 would be worse than the problem -- the check would
 * then measure a real contrast ratio against a colour nobody can see and
 * could fail the page for it.
 *
 * An opacity is the only property in this animation that can hide anything.
 * Without it the contract is not merely careful, it is airtight: there is no
 * frame, in any state, in any check, in which one of these blocks cannot be
 * measured. A 20px settle with the text legible throughout is also the
 * quieter reading of section 1 -- motion confirms, directs or sets a place,
 * and a paragraph that fades up is doing none of the three more than a
 * paragraph that settles does.
 *
 * `translate` rather than `transform`, for the reason MOTION.md section 10
 * now gives: several of these blocks are device frames that carry a transform
 * of their own, and an animation of that property would take it away for as
 * long as it ran. The independent property composes. */
@keyframes page-rise {
  from { translate: 0 var(--motion-rise); }
}
[data-rise][data-shown] {
  animation: page-rise var(--motion-enter) var(--ease-enter);
}