Project overview

This page exercises html-revisor. Click anything to leave a directive on it.

Decided: editing as markdown

The design decisions for scopes and markdown editing. Each links to the cases below that show it.

question answer see
Editing notation Markdown. A1
What can you pick to edit or direct? Any block: a section, a list, one item, a table, one cell, an svg text label. Each edits its whole content as markdown. A1, C1, C6, D6
Inline elements: hover and directives? Per element. a and img: yes. strong: no, it belongs to its block. span: yes, edited as its content, its tag kept; not inside a pre, where spans are highlighting, and not one that fills its parent, which it would hide. B2, D2, E1, E2
Headings # inside a bigger scope; picked alone, only the text. A3
Line breaks and &nbsp; A <br> is a newline. U+00A0 shows as the character. A4, A5
Lists Markdown - lines, nested lists indented. Add and delete items by editing. A list holding paragraphs or tables: pick the inner blocks too. C1, C2, C3
Code blocks A fenced block: backticks for <pre><code>, tildes for a bare <pre>. A pre holding markup (highlighting) is HTML, so its spans survive. C11, D3
Text beside blocks Paragraphs in the enclosing scope's markdown: text, list, text edits as one. Text saved unchanged stays bare. C9, C10, F3
Tables A table with a header row of plain cells is a markdown table; otherwise HTML. Each cell is a scope too. C6, D4
What markdown can't say Stays as inline HTML (kbd, span class, a target). Inside a bigger scope, a block with attributes is HTML too, inner markup and all, since markdown isn't read inside an HTML block. B1, B2, B3, C4, C8
Images ![alt](src); saving takes the other attributes from the original. D2
What isn't text A stand-in: ⟦svg⟧, ⟦script⟧, restored in order from the prior HTML. A literal ⟦ is \⟦. Hidden elements get stand-ins too, so a save keeps them, and inside HTML the markdown writes, a style's blank lines can't end the HTML block. There, a literal ⟦ is &#x27E6;, and a stand-in takes the next of its kind. D1, D5, D7
Writing markdown Rules make it readable; each block and run of phrasing is checked against the parser, and written as HTML where it wouldn't come back the same. The round-trip holds by construction. A2, B4
Unlikely shapes in HTML ⇄ markdown (empty paragraphs, dl, stat tiles) May be ugly. Spend effort on the shapes agents write. C4, E1, F1
Where markdown converts On the server, where scopes.mjs and linkedom run; the client stays small. No live preview.
Parser and writer markdown-it parses. The writer is our own, in scopes.mjs, because it carries the rules in this table; a general HTML-to-markdown library would need each one overridden.

Decided: element identity

Directives, the open box, and the conflict check find elements by ID. A markdown save rebuilds the HTML of its scope, so a save has to say which new element is which old one.

question answer see
Where does an element's ID live? In the file, as data-hrv. An edit that adds or removes elements shifts a tree path or an index; an ID in the file stays put.
What kind of ID? A counter, for short IDs, with its high-water mark in <html data-hrv-next>. Never reused; an ID lasts as long as its element. Copies of one ID all get fresh ones.
Who assigns them? The formatter, on every GET and first thing on every PUT. Agents leave both attributes alone; an element they add gets its ID on the next load.
Saves and conflicts One change per save, checked against the base of its own element. A change elsewhere in the file brings a notice and the save goes through. A conflict means turn-taking broke: the save fails with an error that suggests a reload.
IDs after a markdown save Reconciled against the prior HTML. Siblings pair the way a diff does: the ones that came back the same, then, between those, the most alike with the same tag. An edited element keeps its ID, a moved one keeps its own, and a new one gets a fresh ID from the formatter. A pair's children pair the same way. C1
Directives across a markdown save Each rides to its element's partner, in order. The markdown doesn't show them.
A directive whose element a markdown save deleted? Goes with it.
IDs on a page leaving the tool A cleanup script strips them and the directives, the formatter in reverse, and the commits get squashed.

Text you can edit

Click any block to edit it: the box shows its content as markdown.

This one has bold and emphasis inside; the box shows them as **bold** and *emphasis*. The bold word belongs to its paragraph: click it and you get the paragraph.

Entities survive editing: Fish & chips <tasty> cost $5.

A plain click on this link opens the directive box; ⌘-click navigates.

Lists

A table

A super-basic, well-formed table with no html attributes.

Component Status
Server Done
Client In progress

Elements whose directives sit beside them

  preformatted text
    keeps its   spacing

A card

Cards are wrappers: clicking the margin inside the card opens the directive box for the card itself.

What is a scope?

A scope is any block you can pick: you leave a directive on it, or edit its whole content as markdown. Each case has the same four parts: its name, the question it asks in grey, the example HTML in a dashed box, and a table of the scopes the example would make. Disagree with a row? Leave a directive on it.

Under each example: its HTML (click to open), then a row for each element you could pick, nested inside the element that holds it. The last column is the markdown you would edit if you picked that element. ⟦svg⟧ marks something that isn't text riding inside the markdown.

A. Clear prose

A1. Plain paragraph

The baseline: one scope, edits as plain text.

The formatter moves out of the server and runs on its own.

HTML 
<p>The formatter moves out of the server and runs on its own.</p>
element what markdown can say what you would edit
<p> all markdown 
The formatter moves out of the server and runs on its own.

A2. Paragraph with markdown-able markup

One scope. Markdown: **bold**, *em*, `code`, [link](…).

Run npm test, then read every failure, especially the first one.

HTML 
<p>
  Run <code>npm test</code>, then read <strong>every</strong> failure, <em>especially</em> the <a href="#scopes">first one</a>.
</p>
element what markdown can say what you would edit
<p> all markdown 
Run `npm test`, then read **every** failure, *especially* the [first one](#scopes).
 <a> all markdown 
[first one](#scopes)

A3. Heading with inline markup

A heading is its own scope. Does the markdown include the #, or is that the element's business?

Moving normalize into formatter.mjs

HTML 
<h3>Moving <code>normalize</code> into <em>formatter.mjs</em></h3>
element what markdown can say what you would edit
<h3> all markdown 
Moving `normalize` into *formatter.mjs*

A4. Line breaks

A <br> becomes a trailing backslash or two spaces in markdown. Which round-trips?

Jess Kerr
html-revisor
October 2026

HTML 
<p>Jess Kerr<br>html-revisor<br>October 2026</p>
element what markdown can say what you would edit
<p> all markdown 
Jess Kerr
html-revisor
October 2026

A5. Entities and non-breaking spaces

Markdown has no &nbsp;. Does the field show a real U+00A0, or the entity?

Fish & chips cost $5 <each> — “quoted” and 5 × 3.

HTML 
<p>
  Fish &amp; chips cost&#160;$5 &lt;each&gt; — “quoted” and 5&#160;×&#160;3.
</p>
element what markdown can say what you would edit
<p> all markdown 
Fish & chips cost $5 &lt;each> — “quoted” and 5 × 3.

B. Inline markup markdown can't say

B1. Phrasing elements with no markdown syntax

These pass through as raw HTML in the markdown. Is that readable enough to edit?

Press ⌘+↵ to save. H2O, x2, highlighted, removed added, struck, DDD, a quote, , fine print, underlined.

HTML 
<p>
  Press <kbd>⌘</kbd>+<kbd>↵</kbd> to save. H<sub>2</sub>O, x<sup>2</sup>, <mark>highlighted</mark>, <del>removed</del> <ins>added</ins>, <s>struck</s>, <abbr title="Domain-Driven Design">DDD</abbr>, <q>a quote</q>, <time datetime="2026-10-12">today</time>, <small>fine print</small>, <u>underlined</u>.
</p>
element what markdown can say what you would edit
<p> markdown with HTML 
Press <kbd>⌘</kbd>+<kbd>↵</kbd> to save. H<sub>2</sub>O, x<sup>2</sup>, <mark>highlighted</mark>, <del>removed</del> <ins>added</ins>, <s>struck</s>, <abbr title="Domain-Driven Design">DDD</abbr>, <q>a quote</q>, <time datetime="2026-10-12">today</time>, <small>fine print</small>, <u>underlined</u>.

B2. Attributes on inline elements

Markdown links carry href and title only. Where do class, target, style go?

See the section and a warning-colored word.

HTML 
<p>
  See <a href="#scopes" title="this section" target="_blank" class="ref">the section</a> and a <span style="color: #bc4c00">warning-colored</span> word.
</p>
element what markdown can say what you would edit
<p> markdown with HTML 
See <a href="#scopes" title="this section" target="_blank" class="ref">the section</a> and a <span style="color: #bc4c00">warning-colored</span> word.
 <a.ref> markdown with HTML 
<a href="#scopes" title="this section" target="_blank" class="ref">the section</a>
 <span> all markdown 
warning-colored

B3. Styled span as a component

PLAN.html does this with yak ids. Raw HTML in the markdown, or a token the editor protects?

Split out the formatter stez, then decide scopes a17m.

HTML 
<p>
  Split out the formatter <span class="pill">stez</span>, then decide scopes <span class="pill">a17m</span>.
</p>
element what markdown can say what you would edit
<p> markdown with HTML 
Split out the formatter <span class="pill">stez</span>, then decide scopes <span class="pill">a17m</span>.
 <span.pill> all markdown 
stez
 <span.pill> all markdown 
a17m

B4. b vs strong, i vs em

Markdown has one syntax for each pair. The formatter makes them canonical; does anyone mind?

b and strong; i and em.

HTML 
<p>
  <strong>b</strong> and <strong>strong</strong>; <em>i</em> and <em>em</em>.
</p>
element what markdown can say what you would edit
<p> all markdown 
**b** and **strong**; *i* and *em*.

B5. Footnote reference

A sup around a link: two nested raw tags, or one markdown link inside raw sup?

Linkedom skips the skeleton.1

HTML 
<p>
  Linkedom skips the skeleton.<sup><a href="#fn1">1</a></sup>
</p>
element what markdown can say what you would edit
<p> markdown with HTML 
Linkedom skips the skeleton.<sup>[1](#fn1)</sup>
 <a> all markdown 
[1](#fn1)

C. Containers: one scope or many?

C1. Simple list

One scope (edit as - item lines, add and remove items) or one per <li>?

  • Formatter
  • Scopes with bold
  • Round-trip
HTML 
<ul>
  <li>Formatter</li>
  <li>Scopes with <strong>bold</strong></li>
  <li>Round-trip</li>
</ul>
element what markdown can say what you would edit
<ul> all markdown 
- Formatter
- Scopes with **bold**
- Round-trip
 <li> all markdown 
Formatter
 <li> all markdown 
Scopes with **bold**
 <li> all markdown 
Round-trip

C2. Nested list

The inner list sits inside the first item's text. Where's the boundary?

  • Stage 1
    • move the code
    • add a CLI
  • Stage 2
HTML 
<ul>
  <li>
    Stage 1
    <ul>
      <li>move the code</li>
      <li>add a CLI</li>
    </ul>
  </li>
  <li>Stage 2</li>
</ul>
element what markdown can say what you would edit
<ul> all markdown 
- Stage 1
  - move the code
  - add a CLI
- Stage 2
 <li> all markdown 
Stage 1
- move the code
- add a CLI
  <ul> all markdown 
- move the code
- add a CLI
   <li> all markdown 
move the code
   <li> all markdown 
add a CLI
 <li> all markdown 
Stage 2

C3. List item with paragraphs

A loose list: the li holds blocks, not phrasing. Scopes are the ps, then?

  1. Round-trip.

    Opening and saving must change nothing.

  2. Markdown in the box.

HTML 
<ol start="3">
  <li>
    <p>Round-trip.</p>
    <p>Opening and saving must change nothing.</p>
  </li>
  <li>
    <p>Markdown in the box.</p>
  </li>
</ol>
element what markdown can say what you would edit
<ol> all markdown 
3. Round-trip.

   Opening and saving must change nothing.

4. Markdown in the box.
 <li> all markdown 
Round-trip.

Opening and saving must change nothing.
  <p> all markdown 
Round-trip.
  <p> all markdown 
Opening and saving must change nothing.
 <li> markdown with HTML 
<p>Markdown in the box.</p>
  <p> all markdown 
Markdown in the box.

C4. Definition list

Markdown has no dl. Each dt and dd its own scope?

directive
An @AI request, stored as an HTML comment.
scope
The unit of editing and directives.
HTML 
<dl>
  <dt>directive</dt>
  <dd>An <code>@AI</code> request, stored as an HTML comment.</dd>
  <dt>scope</dt>
  <dd>The unit of editing and directives.</dd>
</dl>
element what markdown can say what you would edit
<dl> markdown with HTML 
<dt>directive</dt>

<dd>An <code>@AI</code> request, stored as an HTML comment.</dd>

<dt>scope</dt>

<dd>The unit of editing and directives.</dd>
 <dt> all markdown 
directive
 <dd> all markdown 
An `@AI` request, stored as an HTML comment.
 <dt> all markdown 
scope
 <dd> all markdown 
The unit of editing and directives.

C5. Blockquote

With paragraphs inside, the paragraphs are scopes. With bare text, the blockquote is.

The curious mind over the inventive mind.

— Clausewitz

A blockquote holding text directly.
HTML 
<blockquote>
  <p>The curious mind over the inventive mind.</p>
  <p>— Clausewitz</p>
</blockquote>
<blockquote>A blockquote holding text directly.</blockquote>
element what markdown can say what you would edit
<blockquote> all markdown 
The curious mind over the inventive mind.

— Clausewitz
 <p> all markdown 
The curious mind over the inventive mind.
 <p> all markdown 
— Clausewitz
<blockquote> all markdown 
A blockquote holding text directly.

C6. Table cells

Each cell a scope? The header row too? A cell holding a list?

Stages
stage status
Formatter in progress, see PLAN
Scopes
  • decide
  • build
HTML 
<table>
  <caption>Stages</caption>
  <tbody>
    <tr>
      <th>stage</th>
      <th>status</th>
    </tr>
    <tr>
      <td>Formatter</td>
      <td>
        <strong>in progress</strong>, see <a href="#scopes">PLAN</a>
      </td>
    </tr>
    <tr>
      <td>Scopes</td>
      <td>
        <ul>
          <li>decide</li>
          <li>build</li>
        </ul>
      </td>
    </tr>
  </tbody>
</table>
element what markdown can say what you would edit
<table> markdown with HTML 
<table>
  <caption>Stages</caption>
  <tbody>
    <tr>
      <th>stage</th>
      <th>status</th>
    </tr>
    <tr>
      <td>Formatter</td>
      <td><strong>in progress</strong>, see <a href="#scopes">PLAN</a></td>
    </tr>
    <tr>
      <td>Scopes</td>
      <td>
        <ul>
          <li>decide</li>
          <li>build</li>
        </ul>
      </td>
    </tr>
  </tbody>
</table>
 <caption> all markdown 
Stages
 <th> all markdown 
stage
 <th> all markdown 
status
 <td> all markdown 
Formatter
 <td> all markdown 
**in progress**, see [PLAN](#scopes)
  <a> all markdown 
[PLAN](#scopes)
 <td> all markdown 
Scopes
 <td> all markdown 
- decide
- build
  <ul> all markdown 
- decide
- build
   <li> all markdown 
decide
   <li> all markdown 
build

C7. Figure with caption

The caption is prose; the figure is a container; the image is nonprose.

diagram yo
The path, as of today.
HTML 
<figure>
  <svg viewBox="0 0 120 40" width="120">
    <rect x="1" y="1" width="118" height="38" rx="6" fill="none" stroke="#8250df" /><text x="60" y="25" text-anchor="middle" font-size="12">diagram yo</text>
  </svg>
  <figcaption>The <em>path</em>, as of today.</figcaption>
</figure>
element what markdown can say what you would edit
<figure> markdown with HTML and ⟦tokens⟧ 
⟦svg⟧

<figcaption>The <em>path</em>, as of today.</figcaption>
 <text> svg text: markdown inside the label 
diagram yo
 <figcaption> all markdown 
The *path*, as of today.

C8. Details and summary

Summary is phrasing; the body is flow. Two scopes?

Why formatter first?

It can take validation steps.

HTML 
<details open>
  <summary>Why <strong>formatter</strong> first?</summary>
  <p>It can take validation steps.</p>
</details>
element what markdown can say what you would edit
<details> markdown with HTML 
<summary>Why <strong>formatter</strong> first?</summary>

It can take validation steps.
 <summary> all markdown 
Why **formatter** first?
 <p> all markdown 
It can take validation steps.

C9. Div holding text directly

No p, but it reads as a paragraph. Same as a p?

A div with text and no paragraph element.
HTML 
<div>A div with <em>text</em> and no paragraph element.</div>
element what markdown can say what you would edit
<div> all markdown 
A div with *text* and no paragraph element.

C10. Text and blocks mixed in one container

Bare text, then a block, then more bare text. Each text run an anonymous scope? Or does the formatter wrap them in p?

Text before the list, with bold.
  • an item
Text after the list.
HTML 
<div>
  Text before the list, <strong>with bold</strong>.
  <ul>
    <li>an item</li>
  </ul>
  Text after the list.
</div>
element what markdown can say what you would edit
<div> all markdown 
Text before the list, **with bold**.

- an item

Text after the list.
 <ul> all markdown 
- an item
  <li> all markdown 
an item

C11. Code block

A scope edited as a fenced block? Spans from syntax highlighting can't survive markdown. 

node formatter.mjs PLAN.html
echo $?
const x = 1;
HTML 
<pre><code>node formatter.mjs PLAN.html
echo $?</code></pre>
<pre><code><span style="color: #cf222e; font-weight: 600">const</span> x = <span style="color: #0550ae">1</span>;</code></pre>
element what markdown can say what you would edit
<pre> all markdown 
```
node formatter.mjs PLAN.html
echo $?
```
<pre> markdown with HTML 
<pre><code><span style="color: #cf222e; font-weight: 600">const</span> x = <span style="color: #0550ae">1</span>;</code></pre>

D. Nonprose inside prose

D1. Inline icon

An svg inside a sentence. A token in the markdown? Raw SVG would swamp the text.

Click the marker to open the directive box.

HTML 
<p>
  Click the
  <svg viewBox="0 0 16 16" width="14" height="14">
    <circle cx="8" cy="8" r="6" fill="#d6336c" />
  </svg>
  marker to open the directive box.
</p>
element what markdown can say what you would edit
<p> markdown with ⟦tokens⟧ 
Click the ⟦svg⟧ marker to open the directive box.

D2. Inline image

Markdown has ![alt](src), but not width or class.

The armadillo armadillo lives in DESIGN.html.

HTML 
<p>
  The armadillo <img src="data:image/svg+xml,%3Csvg xmlns='http://www.w3.org/2000/svg' width='20' height='12'%3E%3Cellipse cx='10' cy='6' rx='9' ry='5' fill='%23998'/%3E%3C/svg%3E" alt="armadillo" width="20"> lives in DESIGN.html.
</p>
element what markdown can say what you would edit
<p> all markdown 
The armadillo ![armadillo](<data:image/svg+xml,%3Csvg xmlns='http://www.w3.org/2000/svg' width='20' height='12'%3E%3Cellipse cx='10' cy='6' rx='9' ry='5' fill='%23998'/%3E%3C/svg%3E>) lives in DESIGN.html.
 <img> all markdown 
![armadillo](<data:image/svg+xml,%3Csvg xmlns='http://www.w3.org/2000/svg' width='20' height='12'%3E%3Cellipse cx='10' cy='6' rx='9' ry='5' fill='%23998'/%3E%3C/svg%3E>)

D3. Code block in a list item

Prose, then nonprose, then prose, all in one li.

  • Run the formatter: 
    node formatter.mjs try-me.html
    then reload the page.
HTML 
<ul>
  <li>
    Run the formatter:
    <pre><code>node formatter.mjs try-me.html</code></pre>
    then reload the page.
  </li>
</ul>
element what markdown can say what you would edit
<ul> all markdown 
- Run the formatter:
  ```
  node formatter.mjs try-me.html
  ```
  then reload the page.
 <li> all markdown 
Run the formatter:
```
node formatter.mjs try-me.html
```
then reload the page.
  <pre> all markdown 
```
node formatter.mjs try-me.html
```

D4. Table in a list item

Same shape as D3, with a bigger nonprose piece.

  • Exit codes:
    0 formatted
    1 can't fix
HTML 
<ul>
  <li>
    Exit codes:
    <table>
      <tbody>
        <tr>
          <td>0</td>
          <td>formatted</td>
        </tr>
        <tr>
          <td>1</td>
          <td>can't fix</td>
        </tr>
      </tbody>
    </table>
  </li>
</ul>
element what markdown can say what you would edit
<ul> markdown with HTML 
- Exit codes:
  <table>
    <tbody>
      <tr>
        <td>0</td>
        <td>formatted</td>
      </tr>
      <tr>
        <td>1</td>
        <td>can't fix</td>
      </tr>
    </tbody>
  </table>
 <li> markdown with HTML 
Exit codes:
<table>
  <tbody>
    <tr>
      <td>0</td>
      <td>formatted</td>
    </tr>
    <tr>
      <td>1</td>
      <td>can't fix</td>
    </tr>
  </tbody>
</table>
  <table> markdown with HTML 
<table>
  <tbody>
    <tr>
      <td>0</td>
      <td>formatted</td>
    </tr>
    <tr>
      <td>1</td>
      <td>can't fix</td>
    </tr>
  </tbody>
</table>
   <td> all markdown 
0
   <td> all markdown 
formatted
   <td> all markdown 
1
   <td> all markdown 
can't fix

D5. A diagram in a paragraph

A big svg between sentences. Does it split the paragraph into two scopes?

Here is the path. It has six stages.

HTML 
<p>
  Here is the path.
  <svg viewBox="0 0 200 30" width="200">
    <line x1="10" y1="15" x2="190" y2="15" stroke="#ccc" stroke-width="4" /><circle cx="10" cy="15" r="6" fill="#1a7f37" /><circle cx="100" cy="15" r="6" fill="#0969da" /><circle cx="190" cy="15" r="6" fill="#0969da" />
  </svg>
  It has six stages.
</p>
element what markdown can say what you would edit
<p> markdown with ⟦tokens⟧ 
Here is the path. ⟦svg⟧ It has six stages.

D6. Text inside an SVG

SVG text takes directives today. Is it a scope too, editable as markdown?

Formatter alone
HTML 
<svg viewBox="0 0 240 40" width="240">
  <text x="10" y="25" font-size="14">Formatter <tspan font-weight="700">alone</tspan></text>
</svg>
element what markdown can say what you would edit
<text> svg text: markdown inside the label 
Formatter <tspan font-weight="700">alone</tspan>

D7. Form controls in text

Unlikely in an overview page, but phrasing content. Nonprose token?

Show done stages, or .

HTML 
<p>
  Show <input type="checkbox" checked> done stages, or <button>reset</button>.
</p>
element what markdown can say what you would edit
<p> markdown with ⟦tokens⟧ 
Show ⟦input⟧ done stages, or ⟦button⟧.

E. Phrasing that isn't prose

E1. Stat tiles

All spans, so phrasing by the spec. But it's layout. Scopes per tile, per span, or none?

6stages
1in progress
HTML 
<div class="stats">
  <div class="stat">
    <span class="big">6</span><span>stages</span>
  </div>
  <div class="stat">
    <span class="big">1</span><span>in progress</span>
  </div>
</div>
element what markdown can say what you would edit
<div.stats> markdown with HTML 
<div class="stat"><span class="big">6</span><span>stages</span></div>

<div class="stat"><span class="big">1</span><span>in progress</span></div>
 <div.stat> markdown with HTML 
<span class="big">6</span><span>stages</span>
  <span.big> all markdown 
6
  <span> all markdown 
stages
 <div.stat> markdown with HTML 
<span class="big">1</span><span>in progress</span>
  <span.big> all markdown 
1
  <span> all markdown 
in progress

E2. A label row

Spans side by side in a flex row. One sentence, or three labels?

stage 1Formatter→ done
HTML 
<div class="row">
  <span class="pill">stage 1</span><span>Formatter</span><span>→ done</span>
</div>
element what markdown can say what you would edit
<div.row> markdown with HTML 
<span class="pill">stage 1</span><span>Formatter</span><span>→ done</span>
 <span.pill> all markdown 
stage 1
 <span> all markdown 
Formatter
 <span> all markdown 
→ done

E3. A link wrapping blocks

A card that's one big link. The link is the container; its heading and paragraph are scopes?

DESIGN.html

Everything the tool adds to a page.
HTML 
<a class="tile" href="#scopes">
  <h4>DESIGN.html</h4>
  <p>Everything the tool adds to a page.</p>
</a>
element what markdown can say what you would edit
<a.tile> all markdown 
#### DESIGN.html

Everything the tool adds to a page.
 <h4> all markdown 
DESIGN.html
 <p> all markdown 
Everything the tool adds to a page.

E4. Paragraph that is all link

Edit the link text, or the paragraph? Same scope either way?

Back to the top of the cases

HTML 
<p><a href="#scopes">Back to the top of the cases</a></p>
element what markdown can say what you would edit
<p> all markdown 
[Back to the top of the cases](#scopes)
 <a> all markdown 
[Back to the top of the cases](#scopes)

F. Odd shapes

F1. Empty and blank paragraphs

Scopes with nothing to edit. Skip them, or let you type into them?

 

Between two empties.

HTML 
<p></p>
<p>&#160;</p>
<p>Between two empties.</p>
element what markdown can say what you would edit
<p> nothing to edit
<p> nothing to edit 
 
<p> all markdown 
Between two empties.

F2. Sibling scopes with no wrapper

A heading and paragraphs straight in a section. Each its own scope, or the section as one markdown doc?

A section

First paragraph.

Second paragraph.

After a rule.

HTML 
<section>
  <h4>A section</h4>
  <p>First paragraph.</p>
  <p>Second paragraph.</p>
  <hr>
  <p>After a rule.</p>
</section>
element what markdown can say what you would edit
<section> all markdown 
#### A section

First paragraph.

Second paragraph.

---

After a rule.
 <h4> all markdown 
A section
 <p> all markdown 
First paragraph.
 <p> all markdown 
Second paragraph.
 <p> all markdown 
After a rule.

F3. A list inside a paragraph

Written as <p>Before the list<ul>…</ul>after the list</p>. The formatter does what a browser does: it closes the p at the ul, leaves the trailing text bare (C10), and turns the stray </p> into an empty p (F1). Leave it, or should the formatter wrap the bare text?

Before the list

  • inside
after the list

HTML 
<p>Before the list</p>
<ul>
  <li>inside</li>
</ul>
<p></p>
element what markdown can say what you would edit
<p> all markdown 
Before the list
<ul> all markdown 
- inside
 <li> all markdown 
inside
<p> nothing to edit