# Markdown Extended specifications

Please note that this document is still a DRAFT of the final specifications: some work remains before the first published version and all its content may change before this publication.

TODOS

• global cleanup & review
• english !! (mistakes and vocabulary)
• the "@TODO" tag shows actual todos

This document describes the specifications of the Markdown Extended syntax. It intends to be a concise and complete set of syntax's rules and tags to use to write under Markdown Extended and a reference to build parsers implementations. It can be considered as the Markdown Extended reference for any purpose. The goal of these specifications is NOT to explain how to write a content but how to build a content using the Markdown Extended syntax.

This document is authored and maintained by Pierre Cassat (@pierowbmstr) and licensed under Creative Commons - CC BY 3.0 ; it is opened for discussion or contribution, please refer to the dedicated section.

## Introduction {#introduction}

Markdown is originally a plain text formatting syntax created by John Gruber and Aaron Swartz. It allows to write some contents with an easy-to-read, easy-to-write set of rules for plain text format then to convert it in a rich format (basically HTML).

Many developers have proposed their own implementation of the original syntax with specific evolutions and extensions. The goal of Markdown Extended is to define an homogeneous new version, as complete and rich as possible, while keeping only the relevant evolutions or more used rules. Markdown Extended can be considered by developers as the new standard for Markdown, so that each user does not need to adapt to the current implementation but can use these rules everywhere.

## Origins of Markdown Extended {#origins-of-markdown-extended}

The original idea of the Markdown syntax came from John Gruber, who defined its goal, the first Markdown syntax rules and coded a first parser as a Perl script. He also wrote the first manual of the syntax.

Working on these specifications, the following implementations inspired us:

We also relied on the Standard Markdown Spec written by John MacFarlane.

For information, developers and webmasters can use the Markdown Mark icon created by Dustin Curtis to identify Markdown files and syntax.

## Scope of these specifications {#scope-of-these-specifications}

The key words "MUST", "MUST NOT", "REQUIRED", "SHALL", "SHALL NOT", "SHOULD", "SHOULD NOT", "RECOMMENDED", "MAY" and "OPTIONAL" in this document are to be interpreted as described in RFC 2119. By extension, any word in capital letters SHOULD be understood in its literal meaning.

The present specifications of the syntax DOES NOT suppose about the final rendering of the content. This rendering is the purpose of the parsers and specific applications (this concept is developed in §§). The "HTML" final output CAN be used as a rendering example but MUST NOT be a specification reference.

Each item of the sections below is identified by a structural ID composed like A.B.C. to allow it to be referenced and used in citations, implementations, documentations etc. These MUST NOT change for a given release of the specifications (current release is indicated at the top of this document).

## Terms and definitions {#terms-and-definitions}

For the purposes of this document, the following terms and definitions apply:

MDE - shortcut for "Markdown Extended".

inline element - An "inline" element is a piece of content that does not concern an entire paragraph ; inline elements are described in the typographic rules of these specifications (§§).

block element - A "block" element is a piece of content that concerns an entire paragraph or more as described at §§ ; block elements are described in the structural rules of these specifications (§§).

blank line - A "blank line" in MDE is any line that does not output anything ; this concept is developed in §§.

meta character - Any character used in at least one rule of the syntax is considered as a "meta" character ; they are listed at §§.

meta data - A "meta-data" is a piece of information written in a document (or loaded by parsers) but not actually rendered in final output ; this concept is developed in §§.

reference - A "reference" is a mark of any kind that refers to a definition written elsewhere in the document (at its bottom for instance) ; this concept is developed in §§.

ID - This designates an identifier string as defined in the HTML language ; it is used in MDE to identify certain contents with a unique (and constant) string ; this concept is developed in §§.

REFID - This designates an identifier string used to identify each reference item ; the REFID follows different construction rules than the classic ID ; this concept is developed in §§.

EOL - For "end of line" ; in an MDE document, this can be either a "CR" \r, "LF" \n or "CRLF" \r\n character depending on the platform ; this concept is developed in §§.

indentation - In MDE like in many other languages, the "indentation" is made by writing a tabulation OR four spaces from last indentation limit (originally the left side of the document, which is the "indentation zero" limit) ; this concept is developed in §§.

reader - In these specifications, "reader" designates the "final user", the persons who read an MDE document.

writer - In these specifications, "writer" designates the "syntax user", the persons who write contents using the MDE syntax.

parser - In these specifications, "parser" designates the application in charge to treat the original MDE content and transform it in a rich language ; the final rendering MUST NOT be important while reading the specifications.

implementation - This designates the same as "parser".

configuration - This designates an optional set of options a parser can accept to define some of its behaviors.

For the purposes of this document, the following notations apply:

∙ - This character is used when necessary to represent a space character.

-> - This notation is used when necessary to represent a tabulation character.

| - This character is used when necessary to represent the left side of the document or the character zero of a line.

## A. Basic concepts {#A}

The sections of this document will explain each tag to use for each writing rule. To begin, as a very first introduction to the MDE syntax, we MUST ALWAYS keep the following basis in mind:

• a Markdown content is written as plain text:
• it MUST be working by any software reading file (such as vi)
• it MUST be readable by a human "as-is" (this is the very first goal of Markdown)
• a paragraph is created in Markdown passing a blank line (this rule is developed in §§)
• as Markdown rules are written using some specific characters, these characters MAY be escaped to be used "as-is" (this is developed in §§)
• all the rules CAN be used in one single Markdown content and be parsed correctly (any conflict between rules MUST be avoided)
• for convenience, the "references" notation MUST be allowed for a maximum of rules (this notation is explained in §§) as it permits to keep a content readable.

### A.1. Intention {#A1}

As said in the scope of these specifications (§§), we WILL NOT give rendering rules here. Each parser MAY follow its own final rendering rules according to the MDE specifications (concerning the original MDE content) and concerned language. Thus, we MUST keep in mind that the final output MAY NOT be HTML only ; a parser can construct any format of output (such as PDF, OpenDocument etc). This is a major difference with the original John Gruber's Markdown parser, which only constructs HTML output.

In other words, we focused on the Markdown document itself:

• the work of the writers who use the syntax to write some rich contents (for various usages listed below)
• the playback quality of the readers who open these contents to read or search an information in them (without transformation).

So the rules we have tried to impose on us by writing these specifications are:

• to allow a writer to construct a content as rich as possible, without being so strict about notations ; some information MUST be writable after OR before a mark, with OR without a space separator etc. to let the writer free to concentrate on the content (more than on the rules themselves)
• to allow a reader to understand quite well the content intentions (the various marks roles) and, most important, to be able to read a content without having a big effort to deploy to ignore the various syntax's marks.

Giving that, we also tried to identify the various usages of the Markdown syntax:

• the first intention of Markdown was to build web contents (let's say a blog article) so a "classic" web page MUST be writable in MDE ; this case also includes input fields and textareas using the Markdown syntax
• it is also commonly used for software documentation (Github uses Markdown as its default "README" files syntax for instance) so MDE MUST easily handle code samples
• Markdown can also be used to write "literary" contents (a dissertation for instance), which means that MDE MUST allow writers to add footnotes, bibliographic and lexicon notes to their contents
• finally, Markdown can also be used for scientific literature and should therefore be able to handle scientific notations (equations, mathematics etc).

With all the lists above, we now have a well-defined scope for these specifications.

### A.2. Global construction {#A2}

The master idea of Markdown is the readability of the content. Which means: if a reader has a Markdown file and opens it with a program like less or vi (or any program that renders a file "as is"), he MUST be able to read its content with no other action.

This idea basically means two things:

1. the semantic tags used to finally build a rich content SHOULD NOT prevent a clear reading of that content
2. a reader SHOULD have to only scroll down to continue reading (right/left scroll SHOULD be avoided)

Keeping readability using the semantic tags of the syntax is the purpose of these specifications ; we will here try to define the best rules to NOT prevent a clear reading but DO allow a maximum of rich final rendering.

Keeping a document "one way scroll only" is simplified by the global construction of a Markdown content that allows the soft breaks of lines. This concept is described at §§.

### A.3. Classification rules {#A3}

The syntax's rules of these specifications are separated in the following four types based on their utility:

Typography : Rules concerning the rendering of typographic writing usages, such as bold emphasis or links ; this almost concerns a word, a group of words or an expression. The typographic rules are often called "inline elements". See §§.

Structure : Rules concerning a special rendering of a sentence or a group of sentences, such as citations or pre-formatted blocks. The structural rules are often called "block elements". See §§.

References : Rules concerning a special feature to make the content light-weight and easier to read by moving some secondary information outside of that content. See §§.

Miscellaneous : Any rule that can not be classified in the three first categories. See §§.

### A.4. Rules precedence {#A4}

The first rule to keep in mind is that the structural construction MUST be parsed BEFORE the typographic one. In other words, blocks rules have precedence over inline ones.

### A.5. Identifiers and indexes {#A5}

An MDE parser MUST be able to reference some specific contents with identifiers (ID and REFID as described at §§). Each identifier must be unique for the document and SHOULD be constant to let readers retrieve a section with permanent links. The identifiers construction process is described in a dedicated section §§.

In some cases, identifiers CAN be defined by the writer to be sure it will be constructed like he expects. Please see dedicated section §§.

### A.6. Retro-compatibility {#A6}

The MDE rules MUST keep ALL original Markdown's rules valid as described on daringfireball.net for retro-compatibility.

## B. Global rules {#B}

### B.1. Application scope {#B1}

This section will describe some rules concerning the whole MDE document and the basis of the syntax.

### B.2. The MDE document file {#B2}

#### B.2.a. File extension {#B2a}

An MDE file MUST be considered as a plain text document. It is basically a raw plain text file just like a classic .txt file. Any MDE file MUST work with any software reading file contents.

Writer Note: The best practice is to always use the same extension for MDE files as it permits to identify them quickly and define some special treatments based on a file extension filtering ; the .md, .mde or .markdown extensions are given here as examples.

#### B.2.b. File encoding {#B2b}

An MDE file MAY be encoded in a "classic" file encoding such as utf-8 or iso-....

#### B.2.c. End-of-file and end-of-line {#B2c}

In an MDE document, the EOL character can be either a "CR" \r, "LF" \n or "CRLF" \r\n depending on the platform. The chosen character MUST NOT change the final rendering.

Implementation Note: This means a parser SHOULD first transform the EOL characters to use the same everywhere and then begin the parsing.

On the same idea, there is no special rule about the EOF character, which may depend on the platform.

### B.3. Blank lines {#B3}

A "blank line" in MDE is any line that does not output anything. A line with only spaces or tabulations MUST be considered as a blank line.

### B.4. Line breaks {#B4}

#### B.4.a. Soft breaks {#B4a}

One of the conditions we posed at §§ is to keep a document "one way scroll only". This is simplified by the global construction of a Markdown content: the real end of lines of the content ARE NOT the final ones. Writers CAN, if they need to, render a final end of line (called "hard break" §§), but the first thing to keep in mind writing with the Markdown syntax is that you DO NOT NEED to write your paragraphs chained on a single line. You SHOULD choose a line-wrapping length (a number of characters a line must not exceed - something like 100 characters in this document) and pass to next line each time your current one goes to this length.

Writer Note: This is generally the case of the common LICENSE files: the text of the license is clearly readable because each line of content does not exceed a short number of characters.

#### B.4.b. Hard breaks {#B4b}

If they require to, writers CAN write a true "hard line break" (which will be rendered as a hard break in final output) following the rule described at §§.

### B.5. Indentation {#B5}

#### B.5.a. Indentation level {#B5a}

The indentation level in MDE is 1 tabulation or four spaces. When a syntax rule embeds characters, these characters are counted to estimate the indentation:

1 tab = 4 spaces = 1 indentation level
1 character + 3 spaces = 4 characters = 1 indentation level
2 characters + 2 spaces = 4 characters = 1 indentation level
...


#### B.5.b. Indentation rules {#B5b}

Any block of content that requires an indentation MAY allow any other syntax's rule to be used in that content, considering the position of the first character of the first indented line as the new left side of its own indentation.

For instance, to create a list sub-item (§§), a writer would write (in the example below, the space character is represented by a ∙):

-∙∙∙list item
∙∙∙∙-∙∙∙list sub-item


The syntax's rules that DO NOT require an indentation can begin at less than 4 spaces from last indentation limit.

For instance, an ATX title (§§) can be indented with 3 spaces if writer feels it would be better (in the example below, the space character is represented by a ∙):

|∙∙∙# title


But if the same title is indented by 4 spaces, it is considered as a code block and not as a title (in the example below, the space character is represented by a ∙):

|∙∙∙∙# title


Implementation Note: an indentation composed of less than the MDE indentation level is often called less_than_a_tab in parsers implementations. It generally means "three spaces or less".

Finally, for rules that requires a character AND an indentation, only the first line MUST actually be indented, subsequent lines of the same block can either be indented or not. This is NOT true when the rule does not include a character (like for pre-formatted blocks - §§).

Writer Note: For a document often read "as is" as plain text, it MAY be a good idea to always indent a block for readability.

#### B.5.c. Breaking indentation {#B5c}

The global rules for blocks sequences are:

• passing 1 blank line MUST close last indented block (indentation level goes down by 1), except if the next block is indented exactly like last one and current block allows paragraphing (this is the case of list items for instance - §§)
• passing 2 blank lines MUST close ALL current indented blocks (indentation level goes back to 0), except in a fenced code block (§§), where the content MUST be rendered "as is" (the two blank lines must therefore be rendered as two blank lines).

Example: A classic example is the case of a code block following a list item. If only one line is passed between both, the code block will be considered as part of the list item's content (and must be indented in consequence by 2: 2 tabulations or 8 spaces). If two blank lines are passed, the block must be considered as a fresh new block (the list item is considered finished).

### B.6. Meta-characters {#B6}

Below is a list of all meta-characters used for semantic rules in MDE (the word in italic is the one we will use in this document):

• \ : the backslash
• . : the dot
• : : the colon
• ! : the exclamation point
• # : the hash mark or "sharp"
• * : the asterisk or "wildcard"
• + : the plus sign
• - : the hyphen or "dash" or minus sign
• _ : the underscore
•  : the backtick quote
• | : the pipe
• ( and ) : the parentheses
• [ and ] : the brackets
• { and } : the curly brackets
• < and > : the angle brackets or "chevrons"

## C. Typographic rules {#C}

A rule is considered as typographic when it only concerns one or more words written inline in a paragraph. As long as it does not concern a full block of text, a rule MUST be considered as typographic.

### C.1. Application scope {#C1}

A typographic rule of the MDE syntax MUST begin and end in the same block of text. It CAN begin at the beginning of the block and end at its end, but it MUST NOT concern more than one block.

For instance, if a writer wants to put parts of two paragraphs in italic emphasis (§§), the example below is valid:

*A first block of content.*

*Another* one ...


While the notation below is NOT valid:

*A first block of content.

Another* one ...


### C.2. Emphasis {#C2}

Bold and italic text emphasis MUST keep simple to use and read. The first idea was kept about these effects allowing two different typefaces:

• the underscore _
• the asterisk *.

The basic rule for emphasis is:

• the concerned typeface characters are surrounding concerned content WITH NO space separator
• an emphasis span MUST allow spaces inside.

#### C.2.a. Emphasis with underscores {#C2a}

Italic text is written surrounded by one character:

• 1 underscore _ on the left and 1 underscore _ on the right WITH NO space separator.

Bold text is written surrounded by two characters:

• 2 underscores __ on the left and 2 underscores __ on the right WITH NO space separator.

Example:

_italic words_
__bold words__


#### C.2.b. Emphasis with asterisks {#C2b}

Italic text is written surrounded by one character:

• 1 asterisk * on the left and 1 asterisk * on the right WITH NO space separator.

Bold text is written surrounded by two characters:

• 2 asterisks ** on the left and 2 asterisks ** on the right WITH NO space separator.

Example:

*italic words*
**bold words**


Writer Note: For a document often read "as is" as plain text, the asterisk notation MAY be used preferably as it seems to let the reader understand word's importance.

#### C.2.c. Emphasis auto-escaping {#C2c}

In-words underscores MUST NOT be considered as emphasis delimiters.

For instance, writing my_underscored_words, the underscored word MUST NOT be in italic but the whole expression MUST be kept "as-is". In other words, writing __my_underscored_words__ MAY be rendered as the expression my_underscored_words written in bold text (leading and trailing underscores are considered as emphasis delimiters but internal underscores are auto-escaped).

This rule is important in a case like underscored_words followed by other_underscores as it seems logical to let a parser match the two distant underscores as delimiters while this MUST NOT be the case (underscores may be kept "as-is" in both expressions).

This rule only concerns underscores. If a writer wants to emphase just a part of a word, he can do so using the asterisks notation.

Example:

in-w*or*d em**phas**is


Implementation Note: As concluded by Fletcher Penney in his work on MultiMarkdown, the original reason of this auto-escaping is that "[underscores] caused too many problems with URL's". Effectively, URLs often use underscores as a word separator and these ones MUST NOT be considered as typographic delimiters.

### C.3. Code and variable names {#C3}

• surround the code or variable name between backticks 
• a code span MUST allow spaces inside.

Example:

This variable \$var can be accessed using the -> get(...) method


Any character of a code span MUST be rendered "as is" in the final output, as the raw litteral written character.

Implementation Notes: In an HTML output rendering, any character must be transformed to its HTML entity equivalent (for instance, the left angle < must be rendered as &lt;) to be finally rendered by the browser as what was written.

A link concerns any kind of hypertext link: a full URL, an in-page link (a hash tag referencing a section of current document) and any other kind of "clickable" span in an HTML content, such as an email address, a phone number etc.

Links MUST allow the references notation. This is described at §§.

Implementation Note: A link MUST be rendered clickable for all output formats when it is possible. When it is not, the address of the link MUST be written in the output to let reader follow it if he wants.
For instance, a PDF rendering MUST embed clickable links, but an OpenDocument, which can not embed clickable interactions, MUST render a link with the original address visible.

Standalone URLs MUST NOT be automatically transformed in links if they are not written following one of the rules below (§§, §§ and §§). An URL written inline in a content with no specific "link" tag is not a link, it is just a raw URL (in an HTML content, it MAY NOT be clickable).

To make an URL (or any kind of text) clickable:

• surround the clickable content between angle brackets < and > WITH NO space separator.

Example:

<http://link.com/query>


Any text written like that MUST be rendered clickable (with a special treatment if required, such as email addresses links - see §§). NO validation process SHOULD be used.

Links can be written inline in the text, separated in two parts:

• first part is the link text between brackets [ and ]
• second part is the link definition, between parenthesis ( and ):
• the URL of the link (relative or absolute)
• an OPTIONAL title wrapped in double-quotes "
• an OPTIONAL list of attributes (§§)
• first and second parts are stuck (with NO space).

Any internal parenthesis ( or ), double-quote " or equal sign = MUST be escaped (§§) inside the second part of the link. Otherwise, the information can be partial as parsers may stop at the wrong character.

Examples:

[a simple link](http://test.com/)

[a link with a title](http://test.com/ "My optional title")

[a link with a title and some attributes](http://test.com/ "My optional title" attribute1=value attribute2=value)


In-page anchors can be accessed using the notation of inline links (§§) replacing the URL by the section hash.

Example:

[my section](#section-id)


See the dedicated section §§ to learn how to define the hash reference of a part.

Some special treatments MAY be applied to special links such as email addresses. As these kinds of links can evolve, there is no list of them here.

Implementation Note: For instance, in an HTML rendering, an email address should be transformed as a mailto:... link.

Links MUST allow user defined attributes (§§) with a special rule:

• there MUST NOT be a space between the closing parenthesis ) of the link definition and the opening curly bracket { of the user defined attributes.

Examples:

[a simple link](http://test.com/){#link-id .link-class}



### C.5. Images {#C5}

#### C.5.a. Inline images {#C5a}

Images can be written inline in the text, separated in two parts:

• first part is the image alternative text between brackets [ and ] preceded by an exclamation point !
• second part is the image definition, between parenthesis ( and ):
• the URL of the image (relative or absolute)
• an OPTIONAL title wrapped in double-quotes "
• an OPTIONAL list of attributes (§§)
• first and second parts are stuck (with NO space).

Any internal parenthesis ( or ), double-quote " or equal sign = MUST be escaped (§§) inside the second part of the link. Otherwise, the information can be partial as parsers may stop at the wrong character.

Examples:

![alt text](http://test.com/data1/images/1.jpg)

![alt text](http://test.com/data1/images/1.jpg "My optional title")

![alt text](http://test.com/data1/images/1.jpg "My optional title" class=myimageclass style="width: 40px;")


#### C.5.b. Images with attributes {#C5b}

Links MUST allow user defined attributes (§§).

Examples:

![alt text](http://test.com/data1/images/1.jpg){#image-id .image-class}

![alt text](http://test.com/data1/images/1.jpg "My optional title"){#image-id .image-class}

![alt text](http://test.com/data1/images/1.jpg "My optional title" class=myimageclass style="width: 40px;"){#image-id .image-class}


Note that there MUST NOT be a space between the closing parenthesis ) of the link definition and the opening curly bracket { of the user defined attributes.

### C.6. Inline mathematics {#C6}

To write inline mathematics:

• surround the content between escaped parenthesis $$ and $$.

Example:

$$...$$
$$\alpha = (t_1 - t_0)/L$$


Mathematics notation is developed in a dedicated section §§.

### C.7. Inline footnotes {#C7}

A footnote can also be written inline like:

Paragraph with a footnote[^Content of the footnote.].


## D. Structural rules {#D}

A rule is considered as structural when it concerns a whole block of content.

### D.1. Application scope {#D1}

A "block" of content is separated from others by:

• the top or bottom sides of the document
• a blank line (§§)

So the rule here is that for any block of content that CAN be written on multiple-lines, it MUST be written with blank lines before AND after. Basically, this is not the case for "ATX" titles (§§) and horizontal rules (§§).

Note: This MAY increase readability of the content as it will separate each block clearly.

Any block of content MUST allow any other syntax's rule to be used and interpreted. This can create some nested indentations as described at §§.

### D.2. Paragraphs {#D2}

To build a paragraph:

• surround the content between blank lines (at least one before and one after - §§).

Example:

Paragraph 1 ...

Paragraph 2 ...


### D.3. Line breaks {#D3}

To insert a line break:

• write two or more spaces at the end of the line and pass to next line.

Example:

Line 1 ...∙∙
Line 2 ...


Note that hard breaks are only allowed on a line with a content. A line with only spaces (even 2 or more) MUST NOT be considered as a hard break but as a blank line (§§).

### D.4. Horizontal rules {#D4}

To insert an horizontal rule:

• write at least three hyphens -, asterisks * or underscores _ alone on a line
• spaces MUST not care.

Example:

***
----
_ _ _


Note: Even if "spaces must not care" in the horizontal rules notations above, writers CAN NOT begin a line with more than 3 spaces as it will be considered as a code block and so not parsed as an horizontal rule.

### D.5. Titles {#D5}

Titles are built using two notations: "ATX" and "Sextet". The HTML h tag have a limit of 6 levels (h1 to h6) but an MDE parser MUST ignore this limit and be prepared for any depth levels as writer requires in a document.

Implementation Note: In an HTML output rendering, when the title level is up to 6, a parser SHOULD write the title's content in a way that it shows this is a title (in the document structure) but without using the h HTML tag.

A title MUST be written on a single line, except the underlining line of the "Sextet" notation (§§).

Any title content MUST be rendered skipping any leading or trailing spaces. This allows writers to add as many space(s) as they require around a title content.

The titles list of a document MAY be accessible to build a "table of contents" quickly. This is developed at §§.

#### D.5.a. ATX: sharps titles {#D5a}

##### D.5.a.1. Basics {#D5a1}

An "ATX" title is written:

• alone on a single line
• preceded by as many hash marks # as the title level
• it MUST be one or more space(s) separator between the hash mark and the title text
• OPTIONALLY, the title can be "closed" using a random number of hash marks at the end of the line.

Examples:

# Title level 1 (HTML tag h1)

### Title level 3 (HTML tag h3)

### Title level 3 (HTML tag h3) ##


The mandatory space between the hash mark and the title's content allows to differentiate a title from the notation below (a commit number at the beginning of a line for instance):

#123456 a text


Implementation Note: The "ATX" structure is taken from the ATX markup.

##### D.5.a.2. ATX titles with attributes {#D5a2}

ATX titles MUST allow user defined attributes (§§).

Examples:

# Title level 1 (HTML tag h1) {#title-id .title-class}
### Title level 3 (HTML tag h3) ## {#title-id .title-class}
##### Title level 5 (HTML tag h5) {#title-id .title-class}#####


As a title MUST be alone on a line, the attributes DOES NOT NEED to be unspaced from the text. The rule here is that any string wrapped between curly brackets { and } MUST be considered as attributes (which means before OR after the final optional hash marks).

#### D.5.b. Setext: underlined titles {#D5b}

##### D.5.b.1. Basics {#D5b1}

The "dash" notation only concerns the two first levels of titles in a document. They are written:

• underlined by equal signs = for the first level
• underlined by hyphens - for the second.
• the underlining line MUST NOT require to be as long as the title text, any number of equals or hyphens MUST work.

Examples:

Title level 1 (HTML tag h1)
=============================

Title level 2 (HTML tag h2)
-----------------------------

Title level 1 (HTML tag h1)
=

Title level 2 (HTML tag h2)
-


Writer Note: If you know that a document will often be read "as-is" as plain text, the Setext notation MAY be chosen preferably to the ATX one as it seems more comprehensive.

Implementation Note: The "sextet" structure is taken from the Setext markup.

##### D.5.b.2. Sextet titles with attributes {#D5b2}

Sextet titles MUST allow user defined attributes (§§).

Examples:

Title level 1 (HTML tag h1) {#title-id .title-class}
======================================================

Title level 2 (HTML tag h2) {#title-id .title-class}
-


As a title MUST be alone on a line, the attributes DOES NOT NEED to be unspaced from the text.

### D.6. Pre-formatted texts {#D6}

In MDE (like in the original Markdown syntax) a "pre-formatted content" is considered as a "pre-formatted code block", which means that, in an HTML output, the result SHOULD be a <pre><code> ... </code></pre> tag rather than a simple <pre> ... </pre>.

#### D.6.a. Simple notation {#D6a}

A pre-formatted block is written as a paragraph:

• begin each line with 4 spaces (1 indentation level).

Example (the pipe of the example below is not included in the notation and represents line's 1st character):

|∙∙∙∙a pre formed content


Unlike indentation rules for blocks (§§), ALL lines of a pre-formatted block MUST be indented as this is the only rule to identify that kind of content.

Following the indentation and blank lines rules (§§), passing two blank lines or more between two code blocks MUST finally render two separated code blocks.

Implementation Note: This means that, in an HTML output, the following MDE content:

|    a pre formed content
|
|
|    another pre formed content


MUST be rendered as:

<pre>a pre formed content</pre>
<pre>another pre formed content</pre>


Any character of a pre-formatted code block MUST be rendered "as is" in the final output, as the raw litteral written character.

Implementation Notes: In an HTML output rendering, any character must be transformed to its HTML entity equivalent (for instance, the left angle < must be rendered as &lt;) to be finally rendered by the browser as what was written.

#### D.6.b. Fenced code blocks {#D6b}

##### D.6.b.1. Basics {#D6b1}

A "fenced" code block can be written:

• surrounded a content between two lines of tildes ~ or backticks  (at least 3).

Example:

~~~
My code here
~~~


Implementation Note: The rendering of such content MUST be the exact same as for "classic" pre-formatted content (§§).

Like specified in the indentation and blank lines rules (§§), passing two blank lines or more inside a fenced code block MUST NOT be considered as blocks separator but as raw blank lines.

Implementation Note: This means that, in an HTML output, the following MDE content:

~~~
a pre formed content

another pre formed content
~~~


MUST be rendered as:

<pre>a pre formed content

another pre formed content</pre>


Any character of a fenced code block MUST be rendered "as is" in the final output, as the raw litteral written character.

Implementation Notes: In an HTML output rendering, any character must be transformed to its HTML entity equivalent (for instance, the left angle < must be rendered as &lt;) to be finally rendered by the browser as what was written.

##### D.6.b.2. Language information {#D6b2}

An information about the language used in the block can be defined by:

• following the first delimiter by the language name (without space).

Example:

html
My HTML code here



Implementation Note: In an HTML implementation, this feature permits to create a language friendly code block, as recommended by the W3C in the HTML5 code element specification.

##### D.6.b.3. Fenced code blocks with attributes {#D6b3}

Fenced code blocks MUST allow user defined attributes (§§).

Examples:

~~~{#block-id .block-class}
My code here
~~~

~~~html{#block-id .block-class}
My code here
~~~


When a language information si defined on the first line of tildes or backticks, the attributes definition MUST be written last (after the language information) and without space.

### D.7. Citations {#D7}

#### D.7.a. Basics {#D7a}

A blockquoted block is written by:

• preceding each line or only the first one of a paragraph by a right angle bracket >.

Example:

> This is my blockquote,
> where we can include **other Markdown** tags ...

> We can also write our blockquotes
without the superior sign on each line, but just at the beginning of the first one.


Once a blockquote has begin (e.g. as long as no blank line is passed), the content MUST be part of it.

Nested blockquotes can be written with consecutive left angle signs rather than indenting each one:

> This is my blockquote
> > This is a nested blockquote


MUST be the same as:

> This is my blockquote
>> This is a nested blockquote


Writer Note: The notation of blockquotes is the "classic" email-style for citations, this is the reason why the indentation between the angle bracket and the content is NOT required.

#### D.7.b. Original URL information {#D7a}

To precise the URL of the original content cited:

• write this URL at the beginning of the first line between parenthesis ( and ).

Example:

> (http://test.com/) This is my blockquote,
> with a content cited from the original "http://test.com" page ...


### D.8. Lists {#D8}

#### D.8.a. List items {#D8a}

A list is made by writing a character (for unordered one) or a number (for ordered one) at the beginning of a line and then writing the content after 1 level of indentation. This means that at least 3 spaces MUST be present after that character:

1 character + 3 spaces = 4 characters = 1 indentation level


As described at §§, the indentation is only required for the first line of a list item:

-   a list item
on two lines


MUST be the same as:

-   a list item
on two lines


Following the indentation level idea, a sub-item is created by adding 1 level of indentation to the last one. It means that, for a list sub-item, the character (originally written at the beginning of the line) MUST be written after 1 level of indentation that represents its parent list item, and then the rule of indentation applies again.

Example:

-∙∙∙a parent list item (first level of indentation)
∙∙∙∙-∙∙∙a sub-item in that parent (second level of indentation)
-∙∙∙continuation of the original list items


#### D.8.b. Unordered lists {#D8b}

An un-ordered list is written by:

• beginning each entry by an asterisk *, a plus sign + or an hyphen - followed by 3 or more spaces (1 indentation level)
• the character used for each item of a list MUST NOT matter
• to create sub-items, just indent the sub-list twice.

Example:

-   first item
*   second item
-   third item

-   first item
*   second item
- first sub-item
* second sub-item
-   third item


#### D.8.c. Ordered lists {#D8c}

An ordered list is written by:

• beginning each entry by a number followed by a period . and 3 or more spaces (1 indentation level)
• the number used for each item of a list MUST NOT matter.

Example:

1.   first item
1.   second item
1.  first sub-item
1.  second sub-item
2.   third item


Writer Note: A special attention MAY be attached when beginning a line with a raw hyphen, asterisk or plus or with a number followed by a point. For instance, to write the raw string 123. My text, the point after the number MUST be escaped to not be parsed as an ordered list item:

123\. My text


### D.9. Terms definitions {#D9}

A definition is written separated in at least two parts:

• the term on the first line, with no leading space or character
• then, the definition beginning on the second line, with a leading colon : followed by at least 3 spaces (1 level of indentation)
• a term can have multiple definitions, separated by a blank line (§§)
• each definition can have multiple paragraphs.

Example:

Apple
:   Pomaceous fruit of plants of the genus Malus in
the family Rosaceae.

Apple
:   Pomaceous fruit of plants of the genus Malus in
the family Rosaceae.

:   Lorem ipsum.

Apple
:   Pomaceous fruit of plants of the genus Malus in
the family Rosaceae.

Lorem ipsum.


### D.10. Tables {#D10}

#### D.10.a. Basics {#D10a}

The basic table syntax is as simple as trying to build a visual table in plain text.

The rules are:

• every table's line is written alone on a single line
• the first lines are the headers of the table (most of the time one single line)
• it is followed by a mandatory separator line of hyphens -
• each line below is a line of the table content
• columns are separated by pipes | and each line may have the same number of pipes
• spacing is not important except for visual facility
• tables can be written without the leading pipes
• finally, the content of the cells can have other MDE's typographic features like emphasis (§§).

Example:

| First Header  | Second Header    |
| ------------- | ---------------- |
| Content Cell  | Content *Cell*   |
| Content Cell  | Content **Cell** |


#### D.10.b. Columns alignment {#D10b}

Alignment in columns can be specified by using colons : in the separators line:

• a colon on the left of a separator's cell means a left-aligned column: :---
• a colon on the right of a separator's cell means a right-aligned column: ---:
• two colons on the left and the right of a separator's cell means a centered column: :--:

Example:

| First Header  | Second Header | Third header |
| ------------- | ------------: | :----------: |
| Content Cell  | **Cell**      | **Cell**     |
| Content Cell  | **Cell**      | **Cell**     |


#### D.10.c. Cell grouping {#D10c}

To build cell groups (one cell over more than one column):

• write as many final pipes as the content must cover columns, without spaces.

Example:

| First Header  | Second Header | Third header |
| ------------- | ------------: | :----------: |
| Content Cell  |  *Long Cell*                ||
| Content Cell  | **Cell**      | **Cell**     |


Implementation Note: Cell grouping MUST be allowed in lines of content AND header line(s).

#### D.10.d. Table caption {#D10d}

Table caption is written:

• between brackets [ and ] just before or just after the table, on a new single line.

Example:

|               | Grouping                    ||
| ------------- | ------------: | :----------: |
| Content Cell  |  *Long Cell*                ||
| Content Cell  | **Cell**      | **Cell**     |
[ my table caption ]


A table ID (as described at §§) can also be defined. Please refer to the dedicated section about attributes §§.

#### D.10.e. Table with multiple bodies {#D10e}

Separate sets of content for a single table are written separating them with one blank line (§§).

Example:

|               | Grouping                    ||
| ------------- | ------------: | :----------: |
| Content Cell  |  *Long Cell*                ||
| Content Cell  | **Cell**      | **Cell**     |

| New section   |   More        |         Data |
| And more      |           And more          ||


Implementation Note: In HTML, the result should be a table with two tbody sections.

#### D.10.f. Table with attributes {#D10f}

Tables MUST allow user defined attributes (§§).

Example:

| First Header  | Second Header | Third header |
| ------------- | ------------: | :----------: |
| Content Cell  |  *Long Cell*                ||
| Content Cell  | **Cell**      | **Cell**     |
[ my table caption ] {#table-id .table-class}


For a table with no caption, the attributes definition is written on a single line just before or after the table (without any blank line).

For a table with a caption line, the attributes definition MUST be written on the same line as the caption. As each of these information are wrapped in different chains, the order or spaces MUST NOT matter.

### D.11. Mathematics blocks {#D11}

Block of mathematics formula is written by:

• surrounding the content between escaped brackets $ and $ (additionally to blank lines).

Example:

$...$
$\Delta = \frac{\partial U^*}{\partial F} = \frac{12F}{Eb} \int_0^L \frac{x^2}{(t_0 + \alpha x)^3} dx$


Mathematics notation is developed in a dedicated section §§.

### D.12. Images blocks {#D12}

Following the same notation as for inline images (§§), an image CAN be written as a single block alone on a line.

Example:

A paragraph ...

![alt text](http://test.com/data1/images/1.jpg "My optional title" class=myimageclass style="width: 40px;")

Another paragraph ...


This notation MUST be used to build the table of figures and tables (§§).

Implementation Note: In an HTML rendering, a block image MAY be rendered as a figure tag, as recommended by the W3C in the HTML5 specifications.

## E. References rules {#E}

### E.1. Application scope {#E1}

The idea of the "references" comes from the first idea of Markdown: let the content be readable "as is".

So, in some cases, it is most relevant to move an information on a single line anywhere in the document (at its bottom for instance) and only insert its REFID (as described at §§) at its final place.

### E.2. Basics {#E2}

The global construction of references is to replace the content by a REFID and write that content on another line anywhere.

The REFID is written between brackets [ and ] in the content. The reference is written on a single line, anywhere in the document (it MUST be skipped from final rendering), with:

• the REFID at the beginning of the line between brackets [ and ]
• a colon : without space after closing bracket
• then the value of the reference after a space

Example:

This is a paragraph with a [referenced link][linkid] ...

...



In the case above, the final rendering MUST be exactly the same as if it was written like:

This is a paragraph with a [referenced link](http://test.com/ "My optional title") ...

...


Links MUST allow the references (§§) notation.

Example:

This is a paragraph with a [referenced link][linkid] ...



This notation can be simplified if the REFID (as described at §§) is the exact same string as the link text.

Example:

This is a paragraph with a link to [Test.com][] ...

[Test.com]: http://test.com/ "My optional title"


Referenced links MUST allow user defined attributes (§§).

Example:

[Test.com]: http://test.com/ "My optional title" {#link-id .link-class}


As the reference definition MUST be written alone on a line, optional spaces CAN separate the reference's content and the attributes definition.

### E.4. Referenced images {#E4}

Images MUST allow the references (§§) notation.

Example:

This is a paragraph with an embedded image ![referenced image][imageid] ...

[imageid]: http://test.com/data1/images/1.jpg "My optional title"


Just like for links (§§), the image reference notation can be simplified if the REFID (as described at §§) is the exact same string as the "alternate" text.

Example:

This is a paragraph with an embedded image ![image 1][] ...

[image 1]: http://test.com/data1/images/1.jpg "My optional title"


Referenced images MUST allow user defined attributes (§§).

Example:

[image 1]: http://test.com/data1/images/1.jpg "My optional title" {#image-id .image-class}


As the reference definition MUST be written alone on a line, optional spaces CAN separate the reference's content and the attributes definition.

### E.5. Abbreviations {#E5}

An abbreviation is written like a reference (§§) with a leading asterisk *:

• the abbreviation term at the beginning of the line between brackets [ and ] and with:
• a leading asterisk * (without space)
• a trailing colon : (without space)
• then the full text of the abbreviation after a space

Example:

*[HTML]: Hyper Text Markup Language


The result SHOULD be informing about the full meaning of the term any time this whole term is found in the content as a whole word.

Example: In the example above (the "HTML" term), in an HTML rendering, the result SHOULD be an abbr tag each time the "HTML" string is found in the content as a full word. Thus, this MUST NOT concern a string like "HTMLize" as it is not strictly "HTML".

### E.6. The special case of notes {#E6}

Footnotes in a text are a way to increase its meaning, external references, some required definitions for the global comprehension, without overcrowding it. They are often used in books and official contents.

We can distinguish three types of footnotes in a content, depending on their intention:

• a simple footnote is a short additional content that would have no place in the content,
• a glossary note is an explanation about a technical term or a glossary entry,
• a bibliographic note, which refers to another book or work, is a hard reference to let the reader find the original source.

Any footnote MUST be callable multiple times in a single documents using the same note REFID, without creating a new footnote entry at each call.

Implementation Note: Parsers MAY allow, in any output format, to let reader access to a footnote text where it appears in the document content. If the footnote text is written elsewhere (at the document bottom for instance), a link MAY be proposed to go back in the content.

#### E.6.a. Footnotes {#E6a}

To create a footnote reference:

• write its REFID like [^ID]:
• a circumflex ^
• the REFID string without space
• between brackets [ and ] in the content
• write the footnote content as a reference (§§).

Example:

Paragraph with a footnote[^myid].

...

[^myid]: Content of the footnote.


A footnote can also be written inline (§§).

#### E.6.b. Glossary notes {#E6b}

A glossary note is most like a definition. It is attached to a specific term and tries to give one or more explanation(s) of it. Glossary notes have to be considered as definitions list, except that they will all be placed like footnotes (at the end of the content for instance).

A glossary footnote is written like a classic footnote but:

• the first line of the note content contains:
• the concerned term
• preceded by the string glossary:
• OPTIONALLY followed by a character between parenthesis ( and ) that will be used to sort the term in an alphabetical lexicon.

Example:

Paragraph with a glossary note[^myid].

...

[^myid]: glossary: Term (x)
Actual content of the glossary footnote.


The differentiation between classic and glossary footnotes allows to build a lexicon of a document referencing only glossary notes (§§).

#### E.6.c. Bibliographic notes {#E6c}

A bibliographic note is a fully referenced external work. This kind of notes is often used in academic or scientific works. The point is that we have to follow some academic rules for bibliographic notes, basically to cite enough information to let the reader find concerned work easily.

A citation note is written like a classic footnote but:

• the REFID of the note is constructed in two parts like [p. XX][#Doe:1991]:
• the page number between brackets [ and ] (this content of this part is OPTIONAL)
• the reference REFID, constructed like classic footnotes (§§) but preceded by a hash mark #
• the note content follows the same rules as for classic footnotes, but the circumflex is replaced by a hash mark #.

Example:

Paragraph with a citation note[p. XX][#Doe:1991].

...

[#Doe:1991]: FirstName LastName (October 5, 1991). *Title of the work*.
edition for instance ... (Web link) Retrieved September 30, 2011.


In a case where the page number is unknown (or not relevant) writer MUST write the first brackets [ and ] with an empty content.

Example:

Paragraph with a citation note[][#Doe:1991].

...

[#Doe:1991]: FirstName LastName (October 5, 1991). *Title of the work*.
edition for instance ... (Web link) Retrieved September 30, 2011.


In a case where writers need to reference a citation without using it in the content, the [not cited] prefix MUST be used and MUST be case insensitive.

Example:

Paragraph with no citation note.

...

[Not Cited][#Doe:1991]: FirstName LastName (October 5, 1991). *Title of the work*.
edition for instance ... (Web link) Retrieved September 30, 2011.


Implementation Note: Parsers MAY have to follow some typographic academic rules for bibliographic notes, such as naming the authors in bold, writing the title of the work in italic etc.

The differentiation between classic and citation footnotes allows to build a bibliographic index of a document referencing only citation notes (§§).

## F. Miscellaneous rules {#F}

@TODO

### F.2. Escaping of meta-characters {#F2}

Escaping a meta-character in MDE only consists in preceding it by a back-slash \.

The following characters MUST be escaped to be rendered as the raw character they are:

• \\ : the backslash itself
• \. : the dot
• \: : the colon
• \! : the exclamation point
• \# : the hash mark
• \* : the asterisk
• \+ : the plus sign
• \- : the hyphen
• \_ : the underscore
•  : the backtick quote
• \| : the pipe
• $$ and $$ : the parentheses
• $ and $ : the brackets
• \{ and \} : the curly brackets

BUT, as certain tags MUST be written using such notation (this is the case of mathematical notations for instance, §§ and §§), they SHOULD be escaped twice if they are used in couples (one opening and one closing character):

• \$$ and \$$ : the parentheses
• \$ and \$ : the brackets

### F.3. Mathematics {#F3}

One of the goal of MDE (and globally the Markdown syntax) is to write scientific documentation. In this idea, MDE parsers MUST be prepared to handle complex mathematical notations.

Inline (§§) and block (§§) mathematical contents (the part inside both wrappers) MUST follow the LaTeX syntax.

As the LaTeX syntax for mathematics is quite complicated, the raw content of such a notation in an MDE document MAY not be easy to read and understand while opening the document "as-is". BUT, it seems to be the best way to handle scientific notations with all its complexity.

Implementation Note: The final rendering of the mathematical notations can be the scope of a third-party application but an MDE parser MUST integrate such third-party natively.

### F.4. Inline HTML {#F4}

On the same idea as for automatic escaping (§§), inline HTML MUST be authorized in an MDE content. BUT, a consequence of the fact that the final rendering MAY NOT be only HTML (§§) is that raw HTML tags MAY NOT be rendered as writer expects in other formats.

#### F.4.a. Use with caution {#F4a}

In conclusion, raw inline HTML MAY be avoided in an MDE content but parsers MUST be prepared to treat such contents.

Implementation Note: Without being a specification rule, we can conclude that parsers SHOULD be prepared to handle HTML tags for any final output. For any other format than HTML, the best practice MAY be to keep the content "as is" skipping any HTML tag (but keeping the text content).

#### F.4.b. Parsing markdown in HTML {#F4b}

If a writer really needs to use an HTML notation in an MDE document, he can "ask" to this content to be parsed adding a markdown="1" attribute.

For instance, the content below will be rendered "as is":

<div>This will be rendered with **raw asterisks**.</div>


While this will rendered as an MDE content (with strong tag instead of asterisks):

<div markdown="1">This will be rendered with **a strong tag**.</div>


HTML comments MUST be an exception as this is the best way to include a comment in an MDE content. All parsers MUST handle HTML comments as comments (in any language).

As a reminder, the HTML comment tag is constructed surrounded between <!-- and --> and can be multi-line.

Example:

<!-- Comment content -->


Implementation Note: The final result of a comment is not really defined as it MAY depend on the rendering format. By default, the best practice SHOULD be to skip comments from final output as a comment may have been written just for not-visible information. Parsers rendering an HTML format CAN keep comments "as is" (present in the output) as they won't be displayed.

### F.5. Automatic escaping {#F5}

As developed by John Gruber in original Markdown's manual, and because the first goal of Markdown was to build HTML rendering, some special characters considered as "meta characters" in HTML MAY be automatically escaped to construct a valid HTML output. Even if we MUST NOT suppose about the final rendering here (§§), we also posed the condition that any original Markdown rule MUST be valid in MDE (§§). So we MUST keep the original auto-escaping of HTML meta characters.

Outside a code span (§§) or a pre-formatted block (§§), and as long as the character is NOT part of an MDE tag, the following characters SHOULD be escaped when rendering an HTML output:

• the ampersand & SHOULD be rendered as &amp;
• the left angle < SHOULD be rendered as &lt;
• the right angle > SHOULD be rendered as &gt;

Implementation Note: Parsers that rendered a non-HTML output SHOULD introduce the raw original character in the final rendering.

### F.6. Meta data {#F6}

#### F.6.a. Basics {#F6a}

Meta-data can be added to an MDE document when necessary ; it can be the case to specify a special title for the document, its author or any kind of "meta" information.

A meta-data is added writing it at the very top of the document (without any blank line from the top) as a var: val pair:

• the meta-data name is written at the beginning of the line, followed by a colon : and a space
• the meta-data content can be multi-line

Example:

author: John Doe


The name of a meta-data MUST be a kind of "slug": a single string without space which can be considered as an identifier.

Multiple meta-data can be written, beginning each item on a new line, and the "true" content of the document MUST begin after passing at least one blank line after meta-data.

#### F.6.b. Meta-data value insertion {#F6b}

Meta-data can also be used to define variables values available in the document. To use a meta-data value, use notation [%var]:

• the meta-data name preceded by a percent sign % between brackets [ and ].

Example:

mymeta: A value for the meta-data

This is the document content, where I can call [%mymeta] value.


### F.7. User defined attributes {#F7}

#### F.7.a. Tags attributes {#F7a}

A special notation allows writers to define some specific attributes, taken from the HTML markup language: some class names and some identifiers.

Writers can use the following notation, placing it near concerned content:

{#id}
{.class-name}
{.class-name #id}
{#id .class-name1 .class-name2}


This notation can be decomposed in:

• one or zero ID string preceded by a hash mark # (without space)
• zero, one or more class names preceded by a period . (without space)
• each attribute separated from others by a space
• the whole string surrounded between curly brackets { and }

The order of class names and ID MUST not matter:

{#id .class-name1 .class-name2}
{.class-name1 #id .class-name2}


Note: The hash mark notation for ids or period notation for class names are very similar to the selectors used by many javascript libraries (such as jQuery or Scriptaculous).

Such notation concerns the following contents:

The usage of a user defined ID CAN be largely used for any type of content in an MDE document. This allows to reference a part of a content for in-page links writing:

This is a paragraph with an anchor{#my-anchor} I can access with a [link](#my-anchor) ...


Note that writing class names on the same model would not have sense as writer can't define concerned span of content. Such a usage of class names in the middle of a content (outside the content types listed above) MUST NOT be allowed and MUST NOT have any effect.

As developed in the identifiers construction section (§§), any ID defined in the content MUST be used instead of automatic one (with, eventually, modifications to fit the "slug" specifications).

Implementation Note: In an HTML rendering, the output MUST be a tag with user defined class names as class attribute and ID as id attribute. In other output formats, some class names SHOULD have an effect in the final rendering (by defining a special rendering for common class names for instance).

#### F.7.b. Raw attributes {#F7b}

In some notations, writers can add raw attributes to contents. This basically concerns only links (§§) and images (§§ and §§).

The global rule for such a notation is that:

• when the attribute value is unique, the simple notation var=val is allowed
• when the attribute value is "long" (includes multiple words), the notation MUST use double-quoted values " : var="val1 val2"

## G. Implementers specifics {#G}

This section will describe some features and rules the MDE's parsers MUST implement (except where noted).

### G.1. Identifiers construction {#G1}

Basically, three types of contents MUST ALWAYS have an identifier:

• the titles MUST have an ID (§§)
• the tables MUST have an ID (§§)
• the figures MUST have an ID (§§)
• the references MUST have a REFID (§§)

In some cases, the ID CAN be defined by the writer to be sure it will be constructed like he expects. Please see dedicated section §§.

#### G.1.a. REFID {#G1a}

As described at §§, we have to differentiate the ID (§§) and the REFID, the references identifier, which does not have the same usage as classic ID and may not follow the same construction rules. The REFID is the string written at the beginning of the line, between brackets [ and ], in the notations listed described at §§.

A reference identifier MUST be constructed, by default, as:

• a string with or without space
• with NO meta-caracter (§§) except when it is part of the identifier rule (see §§ for examples).

#### G.1.b. ID {#G1b}

For the construction of the IDs described at §§, an MDE parser MUST use a constant "slugification" process. Trying to build these specifications, we found three third-parties definitions useful:

In conclusion, in MDE, the construction of IDs MUST follow the rules below:

• a string all lower-case
• containing letters [a-z], digits [0-9], hyphens -, underscores _, colons : and periods .

These rules MUST be applied when constructing an automatic ID AND when a user defined ID occurs (§§). When a writer specify an ID "by hand", parser MUST take that ID instead of creating an automatic one, but the ID MUST be transformed to follow above specifications if it is not the case.

Example: The example below is given as a demonstration. The final "slugify" process is NOT part of these specifications:

### This is a title
// => id=this-is-a-title


### G.2. Automatic indexations {#G2}

Any MDE parser MUST handle a table of contents, extracting at least all titles of a document. Global document's structure MUST be able to use both titles notations (§§) to build a single table of contents.

To do so, any title MUST have a unique ID as reference (like the id attribute of an HTML tag) and this ID MUST follow the specifications at §§.

#### G.2.b. Table of figures and tables {#G2b}

Any MDE parser MUST handle a table of figures and tables of the document, extracting the following objects:

To do so, any table and figure MUST have a unique ID as reference (like the id attribute of an HTML tag) and this ID MUST follow the specifications at §§.

### G.3. Error handling {#G3}

MDE parsers MUST be prepared to handle syntax errors informing the writer about what caused the error and, OPTIONALLY, how to fix it. The error information MAY be a simple phrase with a line number or the tag's type that causes the error.

### G.4. Structural tags {#G4}

A parser SHOULD let the writer choose its rendering structure by allowing him to use structural tags that SHOULD be replaced by corresponding block (mostly indexes) if it is defined and not empty.

These tags are:

• TOC for the table of contents (§§)
• TOF for the table of figures (§§)
• NOTES for the footnotes (§§)
• GLOSSARY for glossary footnotes only (lexicon - §§)
• BIBLIOGRPHY for citations footnotes only (bibliographic index - §§)

The construction of such tag MAY follow these rules:

• the name of the tag (from the list above) in capital letters
• surrounded between percent sign % and curly brackets { and }
• OPTIONAL spaces surrounding the tag name.

Example:

\{% TOC %\}
\{% TOF %\}
\{% NOTES %\}
\{% GLOSSARY %\}
\{%BIBLIOGRPHY%\}


### G.5. User configuration {#G5}

An MDE's parser SHOULD propose any user to define some preferences with a table of options.

### G.6. Various {#G6}

The rules described in this section are NOT part of the specifications, these are just ideas and commonly used features.

#### G.6.a. Other file inclusion {#G6a}

For facility, it seems to be a good idea to allow writers to include external files.

#### G.6.b. Critic markup {#G6b}

As defined in MultiMarkdown, the Critic Markup rules CAN be useful when working on a Markdown document with a team. It allows to keep an information about additions, modifications and deletions in a content.

Implementation Note: When allowing the CriticMarkup syntax, an MDE parser MUST be prepared to handle different states like old and new and finally consider any modification as validated for a final rendering (the new status MUST be the default).

## Contribute {#contribute}

This document and the MDE's specification are opened for any kind of discussion, argumentation, translation and any other modification.

If you'd like to leave feedback, please open an appropriate issue on GitHub. You could first search in existing tickets if your comment does not already exists. If your ticket concerns a specific rule, please include its reference in its title like X.Y.Z. Ticket title ....

The source of this document is maintained under a GIT version-control repository publicly hosted on GitHub. If you want to participate in any way, just fork and edit the original repository and ask for a pull request opening the appropriate issue ticket. GitHub embeds many tools and procedures to allow a quick and simple "fork/pull request" process.

If you want to contribute by translating this document, follow the steps above and create a copy of the original mde-manifest.md file named like mde-manifest.LN.md where LN is your language code following the ISO 639-1 list.

And finally, THANK YOU for being involved ;)