# The md dialect

One hand-rolled parser (`mrlycore::md`) feeds both faces: the site
renders blocks to html, the pixel face lays the same blocks out as
plain text lines. Anything the parser refuses degrades to literal
text - never an error page, never raw html on the wire.

## Accepted

- `#` and `##` headings. Blank lines separate blocks.
- Paragraphs; consecutive lines join with a space.
- `-` bullets and `1.` numbered lists.
- Fenced code blocks (triple backtick), contents escaped verbatim.
- `**bold**`, `*italic*`, and backtick code spans. Nothing is
  parsed inside a span; an unclosed backtick stays literal. ASCII
  math like `a*b*c` belongs in a span or the asterisks italicize.
- Links: absolute http(s) URLs open in a new tab. Relative links
  become slugs the site build resolves against tracked md files -
  a broken link fails the build.
- Images: `![alt](path)` alone on a line, local paths only. The
  build resolves the path against git-tracked files, copies the
  asset into dist at its repo path, and fails on any miss.
- Pipe tables: header row, `|---|` separator (colons tolerated,
  alignment ignored), body rows. Every row piped at both ends,
  same cell count throughout; cells take the full inline pass.
- Blockquotes: `>` lines; consecutive lines merge into one quote.

## Refused

- `###` and deeper headings.
- Raw html - escaped and shown as text.
- Remote images, inline-position images, and any `javascript:` or
  other protocol in a link or image.
- Malformed tables (missing separator, ragged rows, unpiped ends).
- Everything else degrades to the literal text that was typed.