Should GEML have a style layer at all?

[xiongjy2104]

GEML has no styling. .class is defined as a semantic label with "no styling
implied", and there is no raw-HTML escape hatch. But documents do get rendered —
the browser viewer already injects its own CSS — so the question is not whether
appearance exists, only whether the format has anything to say about it.

One constraint seems to come first, because it eliminates options: a style
must never change what the document says. "Fully legible as plain text without
a renderer" is constraint #1, so if a stylesheet can display: none a block,
inject text with content:, or reorder content, the plain text and the rendered
page are two different documents — and the .gemlhistory hash only covers one.

Three shapes we can see:

• Plain CSS beside the document. Zero new syntax, infinite ecosystem. But it
  makes HTML the de-facto target backend, it can change meaning (above), and it
  does nothing for a terminal, PDF or LaTeX renderer.

• A .gemlstyle sidecar in GEML's own syntax, expressing intent rather than
  properties, which each renderer maps into its medium:

  === style {for=".warning"}
  tone = "caution"
  emphasis = "boxed"
  ===
  

  Backend-neutral, still one primitive, and checkable — a for= that matches no
  block is a warning, the same "references can't rot" property the rest of the
  format has. The cost is inventing a vocabulary and answering "why not CSS"
  forever.

• Nothing. Styling stays outside the format; renderers decide, and .class
  is there for whatever tool reads the document model.

A sub-question either way: which file names which? .gemlhistory is found by
filename convention, but a house style is shared by many documents — so probably
the document names its style (style = "house.gemlstyle" in === meta),
reference-checked like any other pointer.

Concretely: could your document's styling be expressed in ten intent words
like tone and emphasis, or do you need pixel control? An answer either way
decides this.