/* motion.css -- the durations. Additive, like type.css: it defines tokens and styles no
   element, so importing it cannot reach existing markup.

   There were three "fast" values in use before this file existed -- .1s, .12s and .15s,
   chosen a rule at a time -- and a --dur-fast that ONLY ReCal defined, at 140ms, while
   every primitive fell back to .12s. The same component animated at two speeds depending
   on which app rendered it, and nothing said which was right.

   THE VALUES ARE ReCal's, not the fallbacks'. 140/240ms were written down in its spec
   ("Motion, house spring, spec directive 4") and sat beside a --spring easing; .12s/.22s
   were typed inline, one transition at a time, and then copied. Between a number somebody
   specified and a number that accumulated, the specified one is the system. Everything
   moves ~17% slower than the fallbacks did, which is one frame at 60Hz.

   A host may still override them; what it may not do is spell a duration out in a rule.
   lint-tokens enforces that.

   fast    a state change on one element: hover, press, focus, a colour
   med     something resizing or revealing: a collapse, a panel
   layout  a LARGE surface moving: the sidebar, the tab riding its edge, the mobile
           sheet. It is its own step because the alternative was those three landing on
           `med` and collapsing 60-80ms faster than they were drawn to -- and because
           two of them had independently picked ~300ms for exactly this job, which is
           evidence rather than taste.
   slow    something crossing the screen
   hint    not motion at all -- how long a one-time hint stays before it fades */
:root {
  --dur-fast: 140ms;
  --dur-med:  240ms;
  --dur-layout: 300ms;
  --dur-slow: 400ms;
  --dur-hint: 3s;
}