R Open Source RecordTHE TOML CHRONICLE
繁體中文Captured 27 Oct 2026
The TOML chronicle

GitHub · #97 / #98 / #99

Chapters +

Reading settings

Preferences stay in this browser. Browser zoom (Ctrl/⌘ +) and pinch zoom remain available.

03 / THE RECORD

Every statement,
in its context.

English editorial paraphrases, organized by GitHub thread and Facebook post. Original names, source IDs, timestamps and links are retained. These are not verbatim translations of all original posts.

Search all records ↗
#97
First proposal

Demand, rejection and the first appeals

A locale specification develops into a debate about tools, contribution methods and conventions.

Closed · Not planned GitHub thread ↗
01
doggy8088 Contributor · Opening post

A proposal for a Traditional Chinese locale

Taiwanese terminology motivates a proposal covering translation, language tags and the website interface.

Taiwan UTC 2026-10-23 14:11:06

doggy8088 notes that the site has a Simplified Chinese cn version but no Traditional Chinese version. Examples such as 檔案 versus 文件 (file), 資料 versus 數據 (data), 雜湊表 versus 哈希表 (hash table), and 剖析 versus 解析 (parsing) illustrate differences in Taiwanese technical usage. The proposal also mentions readers in Hong Kong and Macau. These are the contributor’s statements about demand, rather than survey findings about regional preferences.

The proposed implementation adds a tw directory, the zh-TW language tag, a language name, a homepage, and translations of TOML v1.0.0 and v1.1.0. It proposes translating from English with Taiwanese terminology, rather than merely converting Simplified Chinese characters.

The scope extends beyond prose: mapping directory codes to language tags, localizing release dates and plain-text links, and making navigation, GitHub link labels and the mobile menu use translated strings. Locales without those translations would fall back to English. Generated content would also receive lang attributes.

Acceptance criteria include a successful build producing publish/tw/, continued display of existing locales, and preservation of normative force, especially MUST, SHOULD and MAY. Code examples would generally remain unchanged apart from translated comments.

02
arp242 Project member · Comment

A maintainer characterizes the issue as an LLM prompt

arp242 rejects agent-mediated submissions and closes the issue in the same second.

Taiwan UTC 2026-10-23 14:17:20

Reply contextResponding to the original proposal in #97.

arp242 describes the proposal as an LLM prompt and asks that agents not be used to submit issues and PRs. The comment does not identify individual translation errors or give conditions for revision and resubmission.

GitHub records the reply and closure of #97 as not planned at 22:17:20 on October 23, Taiwan time. Twenty-three seconds later, the same account closes PR #98 without merging it.

The rejection message demonstrably concerns the agent-mediated submission method. This short reply cannot establish which files the maintainer actually read, or that the project opposes every Traditional Chinese translation.

03
doggy8088 Contributor · Comment

The contributor discloses AI assistance and asks for review of the result

He offers accountability and a smaller scope, while asking what standard led to rejection.

Taiwan UTC 2026-10-23 14:32:52

Reply contextResponding to the first rejection and requesting that PR #98 be reopened.

After closure, doggy8088 explains the terminology differences and the purpose of translating from English. He explicitly acknowledges AI assistance with both the translation and the issue, while saying that he is the person submitting and taking responsibility for the work.

He says the PR follows the README instructions for adding a locale, with small additional language-tag changes. Unable to find a policy prohibiting AI assistance, he asks for its basis. He believes the PR was closed without a content review and asks that the contribution itself be judged.

He offers paragraph-by-paragraph comparison with English, welcomes review by native Traditional Chinese speakers and proposes reducing the scope to the homepage and v1.1.0. He asks whether #98 can be reopened or resubmitted, and whether the objection is to AI assistance itself or concrete problems in this PR.

04
doggy8088 Contributor · Comment

A second response on tools, responsibility and review burden

He cites a Linux mailing-list discussion and again offers to reduce the scope.

Taiwan UTC 2026-10-23 15:07:26

Reply contextFollowing the first appeal; no intervening new maintainer reply is visible.

The contributor tries to bring the discussion back to demand: the existing Simplified Chinese version does not fully meet the needs of Traditional Chinese readers in Taiwan and Hong Kong. He reiterates vocabulary differences and frames the translation as filling a gap.

He compares AI with editors, spellcheckers and translation software. He argues for checking correctness, conformity with conventions and the submitter’s willingness to take responsibility. He again offers comparison with English, preservation of examples and native-speaker review.

The comment links a Linux kernel mailing-list message from October 13 as a reference for his view of tools. He acknowledges limited maintainer time and offers to reduce scope, read through all the text himself, or follow contribution guidelines supplied by the project.

He concludes that he would respect a written prohibition on AI assistance; otherwise he wants reconsideration based on content quality. This is a request for clarification and reconsideration, not a maintainer commitment to accept a new proposal.

05
ShikiSuen Discussion participant · Comment

ShikiSuen distinguishes localization from code contributions

He questions rejection based on AI use and later edits the comment to explain withdrawing another response.

Taiwan UTC 2026-10-23 16:02:36Last edited: Taiwan UTC 2026-10-27 02:47:18 · Latest readable version at capture

Reply contextResponding to arp242’s rejection in #97.

ShikiSuen argues that this contribution is localization, that the contributor’s native language is Taiwanese Traditional Chinese, and that he has reviewed the work himself. He says rejecting such a contribution solely as an LLM prompt could be understood as discrimination against users of a particular natural language.

He argues that the appropriateness of LLM use should be considered by task type. Whether the contributor reviewed sufficiently, and whether the rejection constitutes discrimination, are this participant’s claims; this chronicle does not treat them as independently established facts.

The current ending says that he removed another response because better comments now address the situation. This comment was published on October 24 at 00:02:36 and last edited on October 27 at 10:47:18, Taiwan time.

06
d4n1elchen Discussion participant · Comment

d4n1elchen emphasizes the work involved in human review

He rejects equating machine translation with an absence of quality control.

Taiwan UTC 2026-10-25 17:52:07

Reply contextPosted after #99 closed, responding to the argument that readers can obtain machine translations themselves.

d4n1elchen quotes the suggestion that readers can obtain mechanical translations themselves and disputes the idea that this substitutes for localization. Even with LLMs, he argues, machines can miss cultural nuance and human review of a high-quality translation still takes substantial time.

He notes that the translation industry used machine translation before LLMs. The existence of a tool therefore does not erase human effort. He considers it dismissive and unfair to reduce a contribution to a prompt without assessing its result.

He argues that projects rejecting AI assistance should say so in contribution guidelines, rather than leave contributors guessing. This comment advocates review quality and explicit policy; it does not supply a paragraph-by-paragraph review of the two specifications.

07
brlin-tw Discussion participant · Comment

brlin-tw criticizes both the submission and its rejection

Speaking personally, he calls for observing conventions, actionable feedback and de-escalation.

Taiwan UTC 2026-10-25 19:21:32Last edited: Taiwan UTC 2026-10-25 19:32:08 · Latest readable version at capture

Reply contextAddressing the whole incident, including project conventions, Facebook publicity and maintainer feedback.

brlin-tw introduces himself as a member of Taiwan’s free/open-source software localization community, speaking in a personal capacity. He disagrees with doggy8088’s approach and says that silence in a project’s written rules should not be taken as a welcome for AI/LLM-generated issues or PRs.

He cites Russian PR #86, French PR #96 and Portuguese PR #65 as examples that did not require a specification-like issue and kept descriptions to necessary information. He calls the lengthy description in #98 noise that raises doubts about authenticity, and recommends observing existing conventions where explicit rules are absent.

He also criticizes taking the dispute to Facebook as approaching collective pressure. He recommends seeking clarification after a premature closure, and politely asking other maintainers if escalation is necessary, with social media a last resort. This is his assessment of public discussion, not this chronicle’s finding that mobilization occurred.

He also asks arp242 not to close PRs without actionable feedback or reasons, since contributors otherwise cannot understand expectations or revise their work. He credits doggy8088’s intention to contribute and urges both sides to de-escalate and find a mutually acceptable resolution.

The comment discloses that it was drafted on HackMD and proofread by an LLM. This introduces the particular stage at which a tool is used and the human responsibility involved, beyond a simple AI-use/no-AI-use distinction.

08
pcchou Discussion participant · Comment

pcchou asks for a fairer assessment of public discussion

He supports d4n1elchen’s analysis and sees possible benefits in greater awareness of localization.

Taiwan UTC 2026-10-26 02:07:46Last edited: Taiwan UTC 2026-10-26 02:07:56 · Latest readable version at capture

Reply contextSupporting d4n1elchen while offering a different assessment of making the dispute public.

pcchou describes himself as Taiwanese, with at least ten years in Taiwan’s free/open-source software community and personal experience in paid translation. He supports d4n1elchen’s analysis and asks for a fairer view of doggy8088 making the dispute more widely public.

He suggests the resulting visibility could benefit internationalization, localization and open-standards communities by drawing informed observers to these issues. This differs from brlin-tw’s criticism of the publicity strategy.

He does not request a particular merge action or provide measurable publicity outcomes or a review of this translation. The claimed benefit of publicity is therefore presented as his assessment.

#98
Translation PR

The submission and its technical review

Eight changed files, a concrete accessibility issue and a fix absent from the captured PR.

Closed · Unmerged GitHub thread ↗
09
doggy8088 Contributor · Opening post

The PR changes shared templates as well as translations

PR #98 was opened 24 seconds after the issue and changed eight files.

Taiwan UTC 2026-10-23 14:11:30

The PR links #97 as the issue it would resolve. It lists four tw locale files and changes to the formatter, HTML generator and two shared templates. GitHub lists eight changed files, 2,260 added lines and 11 deleted lines. These are diff statistics, not a translation word count.

The contributor says the translation is directly from English, uses Taiwanese terminology and changes only comments in code examples. He reports matching heading, paragraph, code-block and link counts. He also says each paragraph stays on one line to prevent unwanted spaces between CJK text during HTML generation.

The interface changes map tw to zh-TW and add lookups for dates, plain-text versions and navigation labels. Only Traditional Chinese receives translations of the new interface strings; other locales use English fallback text. Language attributes and label substitutions affect templates shared by all locales.

The contributor reports generating the Traditional Chinese homepage, latest version and both specifications with Node.js scripts and webpack. He gives examples of language-menu, release-date and navigation output. He says several independent review passes were performed, but browser tooling could not reach the local server, so generated HTML was inspected instead. No human reviewers are named in that passage.

The final section discusses reversibility and scope of impact, claiming that text output for existing locales remains unchanged. Copilot’s subsequent review concerns language attributes in the shared templates; it does not comprehensively rate the two translated specifications.

10
copilot-pull-request-reviewer[bot] Automated reviewer · Automated review summary

Copilot review summary: language tags on English fallback labels

The automated review raises one accessibility issue and lists eight changed files.

Taiwan UTC 2026-10-23 14:15:23

Copilot recommends changes because English fallback interface labels can inherit non-English lang attributes, causing incorrect screen-reader pronunciation. The summary links a specific inline comment.

It also summarizes the homepage, specification, locale-mapping and template changes, listing eight files. This is an automated change summary and technical review, not native-speaker review or certification of the entire translation’s meaning or readability.

11
Copilot Automated reviewer · Inline discussion

Inline review: page language and interface-string language can differ

The issue concerns English fallbacks in non-English locales, including mobile and language menus.

Taiwan UTC 2026-10-23 14:15:23

Reply contextOn the first line of the shared navigation template; reviewing commit b9bd601.

In the navigation template, Copilot notes that the new lang attribute identifies the page locale, while the lookup can return English for existing non-English locales. An English label inside a container tagged zh-CN, for example, may be pronounced incorrectly by a screen reader. The mobile menu has the same problem.

It recommends retaining the content language while carrying the actual language of each interface string, explicitly marking English fallbacks as English, and marking each language name in the language menu appropriately.

The inline comment and review summary are two representations of one review. Their source objects are recorded separately without treating them as independent quality assessments.

12
doggy8088 Contributor · Inline discussion

The contributor reports a fix; the commit exists but is absent from the PR

The fix follows the original commit, while the captured PR still lists only the original version.

Taiwan UTC 2026-10-23 14:42:29

Reply contextReplying to Copilot’s inline comment; in_reply_to_id is 4360491030.

doggy8088 says commit 83dd824 fixes the issue: labels carry their actual language, English fallbacks use en and Traditional Chinese uses zh-TW. The outer navigation language attribute is removed, the mobile menu receives equivalent treatment and language-menu entries identify their own language. The specification content container retains the page language.

This chronicle separately verified commit 83dd824cc5581e89ef853ca07221ee5b6ca4219d, created on October 23 at 22:42:18, Taiwan time, with the original PR commit b9bd601 as its parent. The fix exists as a commit, beyond a promise in a comment.

At the cutoff, however, #98 still has head b9bd601a7a8a7c2cf9ab5400c489ebbfb3b88b4f and lists only that commit. The record therefore does not establish that the fix was pushed into this PR or approved in a new Copilot review. The reply was posted after the PR had closed.

The API also records an author-submitted review object with no separate summary text, at the same time as this reply. It is included in the coverage inventory under this response, rather than counted as another text-bearing comment.

#99
Renewed proposal

A second request becomes a dispute about trust

A shorter issue, a Facebook post and a statement after closure.

Closed · Locked GitHub thread ↗
13
doggy8088 Contributor · Opening post

A shorter renewed proposal asks whether a contribution is welcome

#99 starts with personal reading experience and offers a smaller initial scope.

Taiwan UTC 2026-10-25 12:48:46

As a reader from Taiwan, doggy8088 says he often consults toml.io and notices Simplified Chinese but no Traditional Chinese. Terms for file and data illustrate that the issue is not merely character conversion: the existing wording may feel unfamiliar to Taiwanese readers.

He says he translated the homepage, v1.0.0 and v1.1.0 from English in his fork and read through them as a native speaker. Code examples are unchanged except for comments. He asks whether maintainers would welcome a PR, perhaps beginning with the homepage and v1.1.0.

He says he understands if maintenance costs rule out another locale. Compared with #97, this description is shorter, omits detailed technical acceptance criteria and does not mention AI in the opening text. Form alone cannot establish an attempt to conceal AI use; the public Facebook post and subsequent statements also matter.

14
ddman Discussion participant · Comment

ddman: localization involves culture and everyday terminology

The first reply supports Traditional Chinese in both English and Chinese.

Taiwan UTC 2026-10-25 13:26:46

Reply contextResponding to the demand for a Traditional Chinese version in #99.

ddman supports the request. He argues that Traditional Chinese readers’ needs extend beyond character conversion or legibility to cultural context and localized technical terminology. Unfamiliar terms create friction in everyday technical reading.

He mentions Hong Kong, parts of Southeast Asia and Traditional Chinese developers worldwide as well as Taiwan. He sees practical value in native documentation and hopes maintainers will consider the contribution. The comment states the same reasons first in English and then in Chinese.

The comment supports the need for a locale. It does not verify the full translation’s quality or propose a specific AI-use policy.

15
arp242 Project member · Comment

A maintainer links Facebook and directly confronts the contributor

The reply and closure of #99 occur in the same second; the thread is locked ten seconds later.

Taiwan UTC 2026-10-25 14:01:36

Reply contextThe responder brings discussion on Facebook back into the issue by linking the public post.

arp242 links a post on Will 保哥的技術交流中心 and repeatedly asks what is wrong with the contributor. The short response gives no technical conditions for adding Traditional Chinese and addresses the submitter directly.

The linked post says the earlier PR was rejected for using Claude Code; the contributor claims it was neither read nor reviewed. He says he again used AI to write an issue in a more human-sounding voice, reviewed it himself and questions why human translation is mandatory. The post links #99.

The record shows #99 closed on October 25 at 22:01:36 and locked by toml-lang at 22:01:46, Taiwan time, allowing collaborators to continue commenting. arp242 posts another reply at 22:18:53 after the lock, consistent with that restriction.

16
arp242 Project member · Comment

The maintainer accepts the locale in principle but rejects the contribution method

He separates language from submission method while using disparaging terms about the contributor and supporters.

Taiwan UTC 2026-10-25 14:18:53Last edited: Taiwan UTC 2026-10-25 14:24:03 · Latest readable version at capture

Reply contextA statement after #99 was closed and locked; subsequently edited.

arp242 explicitly says he does not object to Traditional Chinese or other languages. He describes the earlier issue as a human forwarding AI output, using the term meat proxy, accuses the contributor of stirring up Facebook supporters, and sees the renewed issue as an attempt to disguise the same practice.

He argues that readers wanting a mechanical translation can use free tools or browser features themselves, while toml.io should contain carefully vetted, accurate human translations. This is one of the thread’s more explicit statements about acceptable translation methods.

He uses disparaging terms including AI slop, accuses the contributor and supporters of tolerating dishonesty, and says the incident will make future zh-TW contributions suspect. Those claims about honesty, mobilization and future contributions are his accusations and assessments, not findings established by this chronicle.

The reply brings together tools, quality standards, trust and public discussion. Other participants in #97 subsequently challenge his understanding of cultural differences and human review effort. At the capture cutoff, none of the three threads has reopened and this PR has not been accepted.

R

Keep the context.
Give judgments a record to stand on.

Back to top ↑