A Short History of Markdown

Mauricio Longo — Sun, 05 Apr 2026 02:18:11 GMT

Origins (2004)

Markdown was created by John Gruber, author of the Daring Fireball blog, in collaboration with Aaron Swartz. Gruber — born in 1973 in Philadelphia, a Drexel University CS graduate, and former developer at Bare Bones Software (makers of BBEdit) — announced Markdown on March 15, 2004, in a post titled “Introducing Markdown” on Daring Fireball.

Gruber’s motivation was simple: he was tired of writing raw HTML for his blog. As he put it, “It grew tiresome, and it just felt like I was making work for myself, and I really thought that HTML made it hard to proofread my work.” He wanted to allow authors to “write using an easy-to-read, easy-to-write plain text format, then convert it to structurally valid XHTML (or HTML).”

Aaron Swartz served as Markdown’s “sole beta-tester,” providing substantial feedback on syntax design. Swartz — who at just 14 had co-authored the RSS 1.0 specification and later co-founded Reddit — had created atx in 2002, a structured text format whose #-prefixed headings were adopted directly into Markdown. This is why hash-style headings are still called “atx-style headers” in specs today. Swartz also wrote html2text, a tool for converting HTML back into Markdown.

Design Philosophy

Markdown’s primary design goal was readability. A Markdown-formatted document should be publishable as-is, as plain text, without looking like it’s been tagged or formatted with special instructions.

The single biggest source of inspiration was the format of plain text email. Markdown’s syntax borrowed from conventions people already used in email and Usenet posts: asterisks for emphasis, dashes for lists, and so on.

Other influences included:

setext (c. 1992) — underline-style headers using === and ---
Textile (c. 2002)
reStructuredText (c. 2002)
atx (Aaron Swartz, 2002) — #-prefixed headers

Markdown was conceived as two things at once: a plain text formatting syntax and a software tool (Markdown.pl) that converts that syntax to HTML.

The Original Implementation

The reference implementation was a Perl script of approximately 1,400 lines that processed text through a series of regular expression substitutions, requiring Perl 5.6.0 or later and the Digest::MD5 module.

Gruber initially released it as a GPL-licensed plug-in for Movable Type and Blosxom, as well as a standalone filter for BBEdit.
Markdown 1.0 was released on August 28, 2004. It placed third in Six Apart’s plugin contest for Movable Type 3.0.
Markdown 1.0.1 followed on December 14, 2004, fixing a bug and changing the license from GPL to a BSD-style open source license.
The last-ever update to Markdown.pl was on December 17, 2004. It was never updated again.

The BSD license included a clause that the name “Markdown” could not be used to endorse or promote derived products without specific prior written permission — a clause that would become significant a decade later.

The Fragmentation Problem

Because Gruber’s specification was informal prose rather than a rigorous spec, and because Markdown.pl was abandoned after December 2004 despite known bugs, implementations quickly diverged. A document that rendered one way on GitHub might render differently on Pandoc or another system. One analysis found that a simple list starting mid-sentence could render in up to 15 different ways depending on the parser.

Major implementations that emerged:

PHP Markdown (Michel Fortin, 2004) — PHP port of Markdown.pl, created just days after Gruber’s announcement
PHP Markdown Extra (Michel Fortin, 2005) — Added footnotes, fenced code blocks, tables, definition lists, abbreviations
MultiMarkdown (Fletcher T. Penney, 2005) — Added metadata, LaTeX output, citations, cross-references, math
Pandoc (John MacFarlane, 2006) — Universal document converter with massively extended syntax
Python-Markdown (Community, 2007) — Python implementation with extension system
GitHub Flavored Markdown (GitHub, ~2009) — Added tables, task lists, fenced code with language IDs, strikethrough

Each implementation added features Gruber’s original lacked, but no two agreed on how to handle the ambiguities in the original specification.

The CommonMark Standardization Effort

The Call for Standards (2012)

On October 25, 2012, Jeff Atwood (co-founder of Stack Overflow) published “The Future of Markdown” on his Coding Horror blog, calling on the community to create an official specification and standard test suites.

The next day, a Markdown Community Group was formed at the W3C. Atwood, John MacFarlane (creator of Pandoc, philosophy professor at UC Berkeley), and others emailed Gruber in November 2012 inviting him to participate. They never heard back.

The “Standard Markdown” Naming Dispute (2014)

John MacFarlane led the writing of the specification, drawing on his experience building Markdown parsers in multiple languages — including the first Markdown parser not based on regex substitutions (Pandoc).

On September 3, 2014, Atwood announced the project as “Standard Markdown.” Gruber’s reaction was swift — he called it “infuriating,” demanded the group rename the project, shut down the domain, and apologize. He initially suggested names like “Strict Markdown” or “Pedantic Markdown,” but then indicated that no form of the word “Markdown” was acceptable in the project name.

Gruber had argued that complete standardization would be a mistake: “Different sites (and people) have different needs. No one syntax would make all happy.” The BSD license’s clause about the “Markdown” name gave him legal standing.

On September 5, 2014, the project was renamed first to “Common Markdown,” then to “CommonMark” (one word). Atwood published an apology. Gruber eventually accepted the name.

The Specification

CommonMark provided what had been missing for a decade:

A rigorous, unambiguous specification with formal grammar
Over 600 conformance tests
Reference implementations in C (cmark) and JavaScript (commonmark.js)
The latest spec version is 0.31.2 (January 28, 2024)

Performance was a major advance: cmark is approximately 10,000 times faster than the original Markdown.pl.

The W3C Markdown Community Group was closed on August 13, 2019, but CommonMark development continues independently. As of late 2025, six critical issues remain before a 1.0 release can be declared.

IETF Recognition

In March 2016, the IETF registered the text/markdown media type via RFC 7763, defining it as an informational standard for identifying Markdown documents in MIME-compliant systems, with required UTF-8 charset and optional variant parameters.

Key Variants and Their Contributions

GitHub Flavored Markdown (GFM)

GitHub began using its own Markdown variant around 2009 to render README files in repositories. It formalized GFM in 2011, and in March 2017 released a formal GFM specification built as a strict superset of CommonMark. GFM introduced tables, task lists (- [x]), fenced code blocks with language identifiers, strikethrough (~~text~~), auto-linking, and emoji shortcodes (:smile:). Footnote support was added in 2021. GFM also includes post-processing sanitization to prevent XSS attacks, blocking potentially malicious HTML tags like

The Creative Cyborg: References