This page holds the Merge Conflicts methods. It covers what the series compares, the six labels every claim carries, and the twelve rules every post passes. It also explains right of reply, corrections, and who’s writing. Every post in the series links here. When a count or a claim changes, the log at the bottom says what changed and when.
🗺️ What does the series compare?
Merge Conflicts asks who decides in open source projects, how they decide, and who can review the decision. The name comes from how git presents a merge conflict. A merge conflict happens when two people change the same line and the tool refuses to guess. Someone has to decide, and the series is about that someone.
Season 2 asks the same governance questions of 12 projects. In alphabetical order: Backdrop, the CSS and HTML standards bodies, Drupal, Ghost, Joomla, MariaDB, MySQL, Node.js, OpenJS, PHP, TYPO3 and WordPress.1 Three supplemental cases sit beside them: xz-utils, ClassicPress, and Ruby Central with RubyGems.2 WordPress is one row in every table, sorted like the others, with no special color.
Season 3 reads WordPress’s record through dated documents: handbook pages, announcement posts and policy revisions, each with its date and what it shows. Seasons 4 to 8 turn to how projects work, grow, license, fund and measure themselves.3 A few pieces are first person, about rooms I was in. Each one says so at the top and carries the account label all the way through.
Python, Apache and Kubernetes were in an older plan. They stay out until someone researches them from primary documents.4 If you came for Kubernetes governance, I’m sorry. The research didn’t go there, and I won’t pretend it did.
🎯 What does the series claim, and what cuts against it?
Label: analysis.
Here’s the claim, at the size the evidence supports. For decisions about people, teams, access and the top role, the WordPress record shows no written process and no review outside leadership.5
That’s it. Not “WordPress has no process.”
The counter sits in the same section, as the Merge Conflicts methods require in every post. Team level work in the research sample follows written processes. Plugin and theme review, release squads, minimum PHP changes and security releases all run on rules someone wrote down and published.6 Season 3 shows the documents on both sides, with dates.
One more limit rides with the claim. The incident record was built from consequential decisions, so it centers on the role that made them. A set built from decisions will always point at whoever decided.7 The 2019 to 2021 record shows named release squads, published “go/no-go” lists and decision posts with options.8 A thin stretch in my file isn’t a quiet stretch in the project.
🏷️ Which six evidence labels do the Merge Conflicts methods use?
Every claim in the series carries at least one of six labels. The Merge Conflicts methods take them from the research’s own definitions.9
| Label | What it means |
|---|---|
| verified | Read in a primary governing document, filing, record or official decision, cited with section or passage |
| source-derived | Stated by a reliable secondary source, not yet checked against a primary record |
| account | A named participant’s attributed account. Not a finding |
| analysis | The researcher’s interpretation of cited facts |
| hypothesis | A claim to test. Not a finding |
| unknown | Searched for and not established |
Three rules ride along with the labels. An account is never written as fact, including mine. Unknown means searched for and not found, which isn’t the same as absent. And silence is never an admission: no reply, no record and no statement mean only that.
On the site, each label appears as a word on a chip, and the word carries the meaning. Color is decoration, so the page works in grayscale and in a screen reader.
📊 What do the counts show, as of October 30, 2026?
The WordPress record behind Season 3 holds 86 distinct incidents in 87 files, as of October 30, 2026. It holds 2,417 claims, 2,406 of them public, and 2,840 sources.10 Those counts change as research continues, and the Merge Conflicts methods treat every count as a floor.
Data panel 1: claims and incidents by year
Denominator: 2,406 public claims and 86 incidents, as of October 30, 2026.11
| Period | Claims per year | Incidents (2017 to 2021 only) |
|---|---|---|
| 2003 to 2023 | 18 to 102 | 2017: 3; 2018: 2; 2019: 3; 2020: 3; 2021: 0 |
| 2024 | 323 | |
| 2025 | 243 | |
| 2026 | 505 |
What the panel can’t tell you: a claim’s year is the first date in its row, often the source’s date and not the event’s. 43 public claims have no date. I dropped five rows that parse as 1970 to 1975 or 2057. Seven candidate incidents for 2017 to 2021 wait for incident files and a verification pass.12 So yes, the numbers spike in 2024. No, that doesn’t tell you what caused what.
Data panel 2: labels on public claims
Denominator: 2,406 public claims, as of October 30, 2026; 11 more are private and withheld.13 A claim can carry more than one label, so the counts sum past 2,406.
| Label | Claims |
|---|---|
| verified | 1,879 |
| account | 376 |
| source-derived | 273 |
| unknown | 258 |
| analysis | 202 |
| hypothesis | 10 |
Ten hypotheses out of 2,417 claims. I’d like to call that discipline. It’s mostly that a hypothesis is a claim I haven’t earned yet.
⚖️ Which twelve fairness rules do these methods enforce?
The research set these rules before the first post, and the Merge Conflicts methods treat each one as pass or fail for a draft.14
Rules 1 to 6: what a post is about
- Structures, not people. Each post’s subject is a mechanism: who holds the name, who appoints the conduct body, who hears an appeal. A person appears only in a public role, acting in public, in their own quoted words.
- WordPress is one row. In comparison posts WordPress sits in the same table as 11 other projects, sorted the same way, with no special color or callout.
- Counterevidence sits beside the finding. Every finding names what cuts against it, in the same section, not in a footnote.
- Credit comes first. The WordPress season opens with what WordPress did well when it wrote things down. Three examples: the 2011 to 2012 GPLv3 process, the 2016 versioned plugin guidelines, and elected team reps.
- Say what the record shows, never the motive. When motive is unknown, the post says unknown.
- Evidence labels are visible. Posts mark what a document shows, what is someone’s account, and what is still unknown, using the labels above.
Rules 7 to 12: how the methods treat people, litigation and replies
- Stated reasons first. Before a post describes what someone did, it quotes the reason they gave.
- Litigation is reported in pairs. A ruling for one side appears beside the rulings that went the other way. A ruling at the pleading stage is called one. Where an evidence dispute is unresolved, the post says so and takes no side.
- Right of reply before publishing. Any post that describes a named person’s or company’s action goes to them first, with a stated deadline, and the post says whether they replied.
- Disclosures in the post. Every relationship that bears on a post appears in the first section where it applies. The full list is below.
- No metaphors that assign roles. No landlords, kings or villains. Each post uses the literal name of the mechanism.
- End on questions a reader can use. Each post closes with questions that work for any project, not a verdict on one.
It reads like a code review checklist for a blog. That’s the point. Nobody should merge to main without review, and this series doesn’t get an exception.
📬 How do right of reply and corrections work?
Before a post describes a named person’s or company’s action, they get the draft section and a stated deadline. The post quotes any reply in full and says whether one came.15 A right of reply can’t compel an answer, and silence means only silence. A post can also meet the rule by describing documents and mechanisms without naming anyone, and some posts do.
Under the Merge Conflicts methods, corrections work in the open. Reply from your own site or in each post’s public discussion thread, and your reply appears on the post as a webmention. Prefer a quiet note? The contact form works too. When a correction changes a claim, this page logs what changed and when. The corrected claim stays visible in the post with its history, so nobody has to take my word for what it said before.
The corrections log
| Date | Post | What it said | What it says now | Raised by | Label change |
|---|---|---|---|---|---|
| none yet |
A correction gets a row when it changes a claim, a number, a label or a quote. A typo doesn’t. “Raised by” names a public reply or says “reader, by contact form.” It never names someone who asked not to be named.
🔁 How is the record built, and how do the methods handle private material?
The research lives in a private working vault, and the Merge Conflicts methods begin there. A script builds a research packet for each post from that vault, and the packet is what I draft from. The script writes only rows marked public. Rows marked private, direct messages, private channels and my own recordings never enter a packet, so they never reach a draft.16 That’s a mechanical rule, not a judgment call made at the keyboard.
Each packet carries its build date, and the counts on this methods page carry theirs. Where a footnote on this site says “Internal,” the document isn’t published, and this page carries what it says. The 2026-10-28 research review stated several numbers that later rebuilds showed were too high. Those numbers are superseded and don’t appear in any post.17
Want to check a claim? Each post cites its sources by date and link. Where a source is a page that has since changed, the footnote names the archive capture it was read in. If you find a primary document the record is missing, send it. That’s a merge I’d like to see.
📎 Where do these rules come from?
Four working documents sit behind this page. None of them is published. This page is their public version, and every post cites them by pointing here, so a footnote that says “Internal” resolves on this page.
- Merge Conflicts series rules (2026-10-30). The claim at the size the evidence supports, the evidence labels, the gates a post must clear before it publishes, and the disclosures. On this page: the claim section, the labels section and “Who’s writing this?”
- Merge Conflicts series outline (2026-10-30). The 100 post plan in eight seasons, the twelve fairness rules, and the research each post draws on. On this page: “What does the series compare?” and the fairness rules.
- The research framework and audit (“Open source governance comparison,” 2026-10-30). The label definitions and the second wave audit the claim condenses. On this page: the labels table and the claim.
- The governance dossier (rebuilt 2026-10-30). The incident files, claim ledgers (IDs such as C-XI-041) and source registers behind every count. A claim becomes public when a post cites it with its label and its source. On this page: the counts.
When a post cites one of these, its footnote names the document, the date, and the section of this page that carries the public text. If a post cites a claim ID, the footnote also names the public source behind it, so you can read what the dossier read.
🪪 Who’s writing this?
I work at Pantheon, a WordPress and Drupal host. In 2024 I worked at GoDaddy. Every relationship that bears on this series is one click below.
Who’s writing this, and every relationship that bears on it
- Pantheon, my current employer, is a WordPress and Drupal host; I wrote about my first day at Pantheon. That covers every hosting post and Drupal as a compared project.
- GoDaddy, where I managed its Five for the Future contributors until July 2026; I wrote about what came next after GoDaddy. It applies to Five for the Future posts and 2022 material.
- FAIR, where I sit on the FAIR Package Manager Technical Steering Committee. Any FAIR post carries it.
- The WPOCC, where I’m a co-founding board member. Posts about community nonprofits and funding carry it.
- The WordPress Training Team, where I’m a faculty member and a former team rep. It applies to Training Team posts.
- Questions I asked on stage at WordCamp US 2024 (September 20, 2024), the WordCamp Europe 2025 fireside (June 7, 2025) and WordCamp Canada 2025. Any post that quotes those sessions says so.
- My own words appear in court exhibits in WPEngine v. Automattic: Dkt. 18-28 and 107-4 (my September 20, 2024 WordCamp US question), Dkt. 18-5 (a September 27, 2024 WP Minute conversation) and Dkt. 69-16.18 The series reports that litigation in pairs. I don’t characterize the exhibits.
Everything I say about my own experience in this series is my account. Each post labels it that way, and these methods hold me to it.
🗓️ What has changed on this methods page?
| Date | Change |
|---|---|
| 2026-10-31 | Merge Conflicts methods page created from the series rules and outline. Counts as of 2026-10-30 |
The series starts with the Merge Conflicts series page. Post 1.0 asks what happens when open source governance procedures aren’t written down.
Footnotes (18)
- Courtney Robertson. “Merge Conflicts, series outline.” Internal working document, 2026-10-30. This page is its public version; see Where do these rules come from?. ↩︎
- Courtney Robertson. “Merge Conflicts, series outline.” Internal working document, 2026-10-30. This page is its public version; see Where do these rules come from?. ↩︎
- Courtney Robertson. “Merge Conflicts, series outline.” Internal working document, 2026-10-30. This page is its public version; see Where do these rules come from?. ↩︎
- Courtney Robertson. “Merge Conflicts, series outline.” Internal working document, 2026-10-30. This page is its public version; see Where do these rules come from?. ↩︎
- Courtney Robertson. “Merge Conflicts series rules,” the claim at the size the evidence supports and the selection rule. Internal working document, 2026-10-30, condensing the research audit. This page is its public version; see Where do these rules come from?. ↩︎
- Courtney Robertson. “Merge Conflicts series rules,” the claim at the size the evidence supports and the selection rule. Internal working document, 2026-10-30, condensing the research audit. This page is its public version; see Where do these rules come from?. ↩︎
- Courtney Robertson. “Merge Conflicts series rules,” the claim at the size the evidence supports and the selection rule. Internal working document, 2026-10-30, condensing the research audit. This page is its public version; see Where do these rules come from?. ↩︎
- Courtney Robertson. “Merge Conflicts series rules,” the claim at the size the evidence supports and the selection rule. Internal working document, 2026-10-30, condensing the research audit. This page is its public version; see Where do these rules come from?. ↩︎
- “Open source governance comparison: framework,” the evidence label definitions. Internal working document, 2026-10-30. Reproduced in full on this page; see Where do these rules come from?. ↩︎
- Governance dossier registers and site data, rebuilt 2026-10-30. Internal. The counts on this page are its public figures as of that date; the dossier is described under Where do these rules come from?. ↩︎
- Governance dossier registers and site data, rebuilt 2026-10-30. Internal. The counts on this page are its public figures as of that date; the dossier is described under Where do these rules come from?. ↩︎
- Governance dossier registers and site data, rebuilt 2026-10-30. Internal. The counts on this page are its public figures as of that date; the dossier is described under Where do these rules come from?. ↩︎
- Governance dossier registers and site data, rebuilt 2026-10-30. Internal. The counts on this page are its public figures as of that date; the dossier is described under Where do these rules come from?. ↩︎
- Courtney Robertson. “Merge Conflicts, series outline,” the twelve fairness rules. Internal working document, 2026-10-30. Reproduced in full on this page; see Where do these rules come from?. ↩︎
- Courtney Robertson. “Merge Conflicts, series outline,” the twelve fairness rules. Internal working document, 2026-10-30. Reproduced in full on this page; see Where do these rules come from?. ↩︎
- Governance dossier registers and site data, rebuilt 2026-10-30. Internal. The counts on this page are its public figures as of that date; the dossier is described under Where do these rules come from?. ↩︎
- Governance dossier registers and site data, rebuilt 2026-10-30. Internal. The counts on this page are its public figures as of that date; the dossier is described under Where do these rules come from?. ↩︎
- WPEngine, Inc. v. Automattic Inc., docket entries 18-5, 18-28, 69-16 and 107-4. Public court record. ↩︎