A Traditional
Chinese translation.
An AI dispute.
The TOML chronicle
A request for a zh-TW locale became a debate about tools, quality, trust and the boundaries of collaboration.
Chapters +
Reading settings
Preferences stay in this browser. Browser zoom (Ctrl/⌘ +) and pinch zoom remain available.
The dispute in one minute
At the capture cutoffs: all three GitHub threads are closed, and translation PR #98 is unmerged.
-
The request
On October 23, Will (doggy8088) proposed a Traditional Chinese locale and submitted translations plus shared-interface changes. He later disclosed AI assistance and said he had reviewed the work.
-
The escalation
arp242 first objected to agent-mediated submissions and closed the original issue and PR. The new issue #99 on October 25 was closed and locked after the maintainer linked the Facebook post and confronted the contributor.
-
The disagreement
The contributor wants the result judged on quality; the maintainer expects carefully vetted human translations for the official site. Review evidence, maintenance costs, tone and wording extending suspicion to future zh-TW translations are also disputed.
-
The responses
Facebook contains support as well as criticism about integrity, public shaming and maintenance costs. In one reply, Will explicitly accepts criticism and acknowledges insufficient consideration; elsewhere he continues to prioritize result quality.
Public statements and recorded actions can be checked. This chronicle has not comprehensively reviewed the translations or resolved every allegation of deception, mobilization or discrimination.
Before the dispute,
there was a contribution.
Demand, submitted work and the public response are distinct parts of this story.
TOML is a text format for configuration files. Its website provides specifications in several languages. On October 23, 2026, doggy8088 (Will, whose Facebook page is named Will 保哥的技術交流中心) proposed a Taiwanese Traditional Chinese locale and submitted a homepage, two specification versions and interface changes. TOML website ↗ Read the proposal
Project member arp242 characterized the first issue as an LLM prompt and objected to agent-mediated submissions. The issue and PR were closed. The contributor disclosed AI assistance, said he reviewed the work and accepted responsibility, and asked for assessment of the result. Two days later, a renewed issue was closed and locked after the maintainer linked Facebook. First rejection · The renewed closure
Responses do not form a single camp: some emphasize human review and cultural differences, some criticize the submission or publicity strategy, and others also ask maintainers for better feedback. This chronicle paraphrases each statement and distinguishes participant claims from facts the record can establish.
Three layers to assess separately
Traditional Chinese localization
Characters, Taiwanese technical terms and reading habits: whose needs support another locale?
Translation and code
Is the prose accurate? Are shared templates correct? What review evidence can others verify?
Rules, effort and trust
Which methods are welcome? How is rejection explained? How does public discussion affect trust?
People in the record
Issue and PR author; publicly discloses AI assistance and accepts responsibility.
Shown as a project member on GitHub; closes threads and explains his position.
Discuss localization, human review and the rejection’s handling.
Describe localization or community experience and differ on publicity and handling.
An automated reviewer; the concrete finding concerns interface language tags.
Who is arp242? Maintenance work and earlier positions
A developer with a long record in Go and technical communities
arp242 publicly identifies as Martin Tournoij; his GitHub profile and personal website link to each other. His public CV describes him as the founder of GoatCounter, an open-source web analytics service focused on privacy and ease of use. NLnet separately records funding for the project between 2019 and 2022. GitHub profile ↗ · His CV ↗ · NLnet funding record ↗
His public work also includes the Unicode lookup tool uni, the Go plugin gopher.vim and maintenance activity on Go libraries including fsnotify and lib/pq. His 2019 resignation post records that he served as a moderator on Vi and Vim Stack Exchange. His project list ↗ · Moderator resignation ↗
Visible work in TOML
arp242 published the TOML 1.1.0 GitHub release, merged the Simplified Chinese and French 1.1.0 translations on toml.io and proposed a CODEOWNERS file to identify translation reviewers. That proposal remained unmerged when checked. These actions establish a maintenance role, but the public record alone does not show whether other members were consulted about the present decision. GitHub release ↗ · Chinese #92 ↗ · French #96 ↗ · Reviewer proposal #89 ↗
His AI contribution policies in other projects
The written bans below are published in other repositories. In toml-lang/toml.io, the repository involved in this dispute, the PR-base README contained no AI-use restriction. In the event discussions, arp242 explicitly asked contributors not to use agents to submit issues or PRs and called for carefully vetted human translations. toml.io README at the time ↗ · Submission restriction in #97 ↗ · Translation requirement in #99 ↗
On August 19, 2026, Taiwan time, he added AI contribution policies to BurntSushi/toml, lib/pq and GoatCounter. They prohibit AI-generated PRs, issues, commit messages and comments. The restrictions cover several forms of contribution text; they cannot be reduced to a ban only on drafts lacking human review. Fixed BurntSushi/toml policy ↗ · Fixed lib/pq policy ↗ · Fixed GoatCounter policy ↗
That day, he described roughly 110 agent reports and a large PR on lib/pq as difficult to review; the count and attribution are his account. The Go vulnerability database’s rolandshoemaker subsequently acknowledged that some reports had slipped through review, said they had been withdrawn and apologized. arp242 also said some issues could still be fixed. This gives context for his discussion of maintenance workload without establishing that all reports were fabricated. His account of review workload ↗ · Withdrawal and apology ↗ · His reply on possible fixes ↗
The fsnotify dispute over maintenance access
In May 2026, removing some collaborators’ access to fsnotify prompted public discussion. arp242 cited rapidly merged changes, inadequate quality and an undiscussed funding-list change. mattn listed fixes and releases, disputed the wholesale dismissal of his work and apologized for incorrectly describing the original maintainers as recently removed and for his judgment in directly changing the funding file. arp242 also acknowledged that he should have opened an issue beforehand. Quality and governance remained disputed; this provides context about maintenance procedures, not proof of his motive in rejecting the translation. arp242’s explanation ↗ · mattn’s full response ↗ · arp242 on prior communication ↗
Earlier views and the acceptance criteria still unresolved
In “Open source DIY ethics,” he describes open source as working together to solve problems and expects users to share the work. In “Tired of Stack Overflow,” he criticizes harsh treatment of newcomers and argues that quality and kindness can coexist. These are his earlier published views, not established explanations for his present replies. Open source DIY essay ↗ · Stack Overflow essay ↗
The PR-base README provides translation steps but contains no AI clause. Policies in other projects do not automatically establish a previously announced toml.io rule. In the Russian discussion, pradyunsg described a practice of finding another native speaker to review. In #99, arp242 said he did not oppose Traditional Chinese, yet also said subsequent zh-TW submissions would be suspected of AI translation. Whether human checking and independent native-speaker review offer an acceptable path remains unanswered in the public record. Translation guide at the time ↗ · Native-speaker review practice ↗ · His full position in this dispute
Background checked on October 27, 2026, Taiwan time. CV details remain attributed to him. This research did not contact participants or recapture all event comments. The background is separate from the counts of 144 records and 28 timeline entries. Source scope and method
Links from the original edition
Sections and records from the original single page now have the following homes. Follow a link to read the full content.
- Three threads.One timeline.
- All records and search
- All records and search
- All records and search
- All records and search
- Several questionsinside one dispute.
- What can be checked,and where the record ends.
- A response no longer present in the current inventory
- Edits, missing comments and snapshot differences
- Russian comparison and the BBC translation workflow
- arp242’s public professional background and maintenance policies
- Three original Facebook share links
- Three original Facebook share links
- The renewed issue and the argument about AI tools
- Reporting full proofreading and publishing a separate Traditional Chinese site
- AI labels and the experience of contributing
- Russian review, tools and locale labels: what does the wording support?
- Russian review, tools and locale labels: what does the wording support?
- Does saying he cannot check Russian imply he speaks Chinese?
- Is rejecting an AI translation while suggesting readers use machine translation contradictory?
- What kind of generalization extends one dispute to future zh-TW translations?
- What does the BBC’s AI-translation disclosure demonstrate?
- #97 Demand, rejection and the first appeals
- #98 The submission and its technical review
- #99 A second request becomes a dispute about trust
- A proposal for a Traditional Chinese locale
- The PR changes shared templates as well as translations
- A maintainer characterizes the issue as an LLM prompt
- ShikiSuen distinguishes localization from code contributions
- d4n1elchen emphasizes the work involved in human review
- brlin-tw criticizes both the submission and its rejection
- pcchou asks for a fairer assessment of public discussion
- A shorter renewed proposal asks whether a contribution is welcome
- ddman: localization involves culture and everyday terminology
- A maintainer links Facebook and directly confronts the contributor
- The maintainer accepts the locale in principle but rejects the contribution method
- Copilot review summary: language tags on English fallback labels
- Inline review: page language and interface-string language can differ
- The contributor reports a fix; the commit exists but is absent from the PR
- The original post’s comments
- Support, criticism, questions and replies
- The renewed issue is closed too
- The author says he cannot understand the outcome
- Comparing the maintainer’s attitude with terminology disputes in Taiwan
- Criticizing the publicity strategy and fearing consequences for zh-TW
- Suggesting Chinese submissions that readers can translate themselves
- The author speculates that he may have been blocked
- Suggesting a challenge about racism
- The author says the issue is closed and some people cannot change
- A sarcastic “generated by a human” signature
- An analogy with a conservative craft workshop
- A brief negative judgment about the interaction
- Questioning why zh-CN appears in a Traditional Chinese PR
- The author explains that the tags support language switching
- Mocking the questioner’s understanding of the code
- A disparaging reply to Logos
- Questioning the basis for “AI slop” while noting shared-code changes
- Speculating about reactions to rapid AI translation and Facebook
- Seeing generated-writing signals in the PR description’s length
- The separately published translation post
- Tools, integrity, maintenance costs and concessions
- The author links his separately published Traditional Chinese site
- Distinguishing the tool debate from integrity
- Starting with maintenance costs and choosing to accommodate maintainers
- Connecting the maintainer’s anger with the previous post’s tone
- Participation means accepting a project’s tool restrictions
- Citing Forgejo’s position on AI submissions
- Alleging broken support for existing locales
- Responding to the author’s anger with emojis
- Finding human and AI writing harder to distinguish
- Explaining open-source caution through low-quality code
- A brush-writing analogy mocking tool restrictions
- Sharing a personal experience of rejected Chinese translation
- Predicting different futures for projects that support or oppose AI
- A Taiwanese-language criticism of language barriers and broader changes
- Approving the term meat proxy
- Favoring doing it yourself and reporting confirmed bugs
- A brief suggestion to ignore the other party
- Responding with a distinction between Taiwan and China
- Reading the incident as an example of opposition to AI
- Suggesting several Taiwanese readers comment to attest to readability
- Questioning how the maintainer identified AI translation
- Speculating with political labels and possible regional misidentification
- Replying with another political generalization
- Generalizing about the number of anti-AI people
- Describing personal complaints about rules and security
- Offering sympathy and good wishes
- Recommending the maintainer’s reply and other readers’ advice
- Rendering the long reply from the locked issue in Chinese
- Attachment 1: interface-label dictionaries and language-code mappings
- Sharing a social-media repost about Linus
- Paraphrasing Linus on people who reject AI
- Attachment 2: language and label changes in navigation
- Attachment 3: specification container and menu-label changes
- Asking for respect for the website maintainer
- Citing Copilot’s advice on English fallback labels
- Attachment 4: brlin-tw’s criticism and call to de-escalate
- Criticizing an echo chamber and equating disagreement with anti-AI views
- The author accepts criticism and acknowledges insufficient consideration
- The author again asks for explicit contribution guidelines
- Connecting the GitHub profile with the Vibe Coder label
- Deriving decision rights from responsibility
- Asking whether a profile label justifies discriminatory treatment
- Acknowledging a misread username and a correction
- Calling for an AI policy agreed by the maintenance team
- Pointing out arp being written as alp
- The author continues to prioritize results over tools
- Explaining rejection through low-quality output overwhelming review
- The author refers to unspecified well-known examples
- The author renders the Taiwanese criticism into Mandarin
- Citing suspicion toward future zh-TW work and doubting contribution prospects
- Emphasizing that the changes go beyond i18n
- The later reflection post
- Public discussion, rejection rights and contribution guidelines
- Questioning whether public discussion amounts to public shaming
- Suggesting a brief agreement to end an argument
- Recognizing years of sharing and community work
- Focusing on understanding and checking AI output
- Connecting the post’s advice with personal difficulties
- Joking about course sales
- Mentioning a fork as an option when people disagree
- Using vulgar language to criticize mockers
- Supporting the protection of personal enthusiasm
- The author admits not expecting the maintainer to see Facebook
- Characterizing the post as venting about an experience
- Opposing an automatic label of public shaming
- Allowing both public discussion and maintainer discretion
- Calling rejected PRs a normal part of open source
- Affirming the intention to contribute and encouraging continued work
- The author says the surprise was being called AI slop
- Locating the complaint in the tone of rejection
- Recommending human editing of generated interactions
- Requesting an advance policy and criticizing the rejection’s tone
- Emphasizing the right to reject and respect for stated requirements
- Acknowledging the right to reject while asking for fuller context
- Agreeing with the brief-response strategy
- Quoting a classical saying about perseverance
- A Buddhist image of steadiness amid outside judgment
- A reader from Macau thanks the author
- Expressing support through a recent class experience
- Distinguishing a sound action from the way it is presented
- Hoping for future articles about Linux
- Recalling learning about development tools through the author’s teaching
- Saying course enrollment is a form of support
- Comparing published work with a family heirloom
- Prioritizing personal enjoyment
- Taking useful criticism while treating the rest as noise
- Sharing experience of PRs left unanswered
- Seeing shortcomings in old work as evidence of growth
- Encouragement to keep going
- Learning new information from the author’s posts
- Praising openness and sustained contributions
- A brief thank-you for the contributions
- A brief “come on” remark with unclear tone
- Encouraging persistence in what one believes is right
- Crediting development experience with better use of AI
- Asking about a live class
- Recalling helpful solutions as a beginner
- Responding with a cartoon GIF containing “good”
- Recommending time for worthwhile things
- Criticizing both a sweeping rejection and a mixed-scope PR
- Questioning the other party’s ability to recognize AI writing
- Comparing the reaction with insults dating back to BBS culture
- A disparaging remark aimed at critics
- Finding the articles useful
- Saying people already know the author
- Valuing explanations in plain language
- Valuing the willingness to share
- An analogy with good gaming performance being mistaken for cheating
- The author jokes that descendants may not appreciate his work
- Leaving the next generation’s opinion to that generation
- The author supplies information about a live class
- Again separating rejection rights, tone and contribution guidelines