Lumen Docs 開源文件平台 · 規劃輸入Lumen Docs open documentation platform · planning input
這份報告不是社群年報，而是明年度翻譯制度、審閱流程、發行節奏三項規劃工作的資料基礎。所有數字皆為示範用途。This is not a community annual report. It is the data behind three pieces of next year’s planning: the translation policy, the review process, and the release rhythm. Every figure here is illustrative.
報告定位WHAT THIS BRIEF IS
社群目前的痛點多半被描述成「翻譯落後」。但把資料攤開之後會發現，落後的不是翻譯本身，而是原文改版之後沒有人知道哪幾頁失效了。The community usually describes the pain as “translation is behind”. Laid out, the data says something else: what is behind is not the translating but the fact that nobody learns which pages went stale when the source changed.
| 觀察到的現象What was observed | 對制度的意義What it means for policy | 頁數Pages |
|---|---|---|
| 首次翻譯完成後未再更新Translated once, never updated since | 缺少原文變更通知機制No notification when the source changes | 180 |
| 翻譯與原文版本相差兩個大版Two major versions behind the source | 讀者看到的是舊行為Readers are shown behaviour that no longer exists | 64 |
| 審閱者僅一人的語系Locales with a single reviewer | 單點依賴，請假即停擺One point of failure; one holiday stops the locale | 3 |
| 與原文同版發行的語系Locales released in step with the source | 流程已經成立，可作為範本The process already works here; copy it | 2 |
唯一需要立刻處理的是最後兩列：一個是風險，一個是可以複製的做法。Only the last two rows need acting on now: one is a risk, the other is a practice worth copying.
本報告的三個章節THE THREE CHAPTERS
章節 01 · 供 翻譯制度規劃 使用CHAPTER 01 · FEEDS THE TRANSLATION POLICY
一頁翻過就永遠算數，是目前所有統計最大的問題。Counting a page forever because it was translated once is the flaw underneath every current statistic.
兩種算法TWO WAYS TO COUNT
兩個數字都沒有算錯，差別在於承不承認文件會改版。
註：本頁所有數字為版型示範用，並非任何真實專案的統計。Neither number is miscalculated. The difference is whether you admit that documentation changes.
Note: every figure on this page is illustrative, and describes no real project.
失效的速度HOW FAST IT GOES STALE
每完成一頁新翻譯的同時，平均有 3.4 頁既有翻譯因為原文改版而失去同步。示範數字，用來說明趨勢而非精確值。For every page newly translated, 3.4 already-translated pages fall out of sync because the source moved. An illustrative figure, meant to show direction rather than precision.
這個比例說明的是方向：只靠增加譯者，覆蓋率會上升，同步率不會。要改的是通知機制，不是人力配置。The ratio points one way: adding translators raises coverage and leaves sync where it was. What needs fixing is the notification, not the headcount.
審閱人力分布WHERE THE REVIEWERS ARE
最後一列不是「人少」，是沒有備援。審閱者一旦停止參與，該語系的發行就停在原地，而且沒有任何流程會發現這件事。The last row is not “short-staffed”, it is unbacked. If that reviewer stops, the locale stops releasing, and no process in place would notice.
寫進明年度規範FOR NEXT YEAR’S POLICY
待社群決定FOR THE COMMUNITY TO DECIDE
下一步NEXT
標記機制一週可以做完，制度討論需要一個月。
兩件事可以同時開始，順序不能顛倒。The marker takes a week to build; the policy conversation takes a month.
They can start together — but not in the other order.