Re: [PATCH v2 16/16] docs/zh_TW: Add a glossary for Traditional Chinese translations

[葉宸佑]

On Tue, Jul 28, 2026 at 09:36:24PM +0800, Alex Shi wrote:
> > +下表同時列出常見的簡體中文（zh_CN）譯法，方便從zh_CN翻譯轉換或對照時
> > +使用。
>
> This is related with copyright. At last, you need to ask the original
> author's permission, and give a reference here. I guess...

Thanks for raising this -- it turned out to be a more useful question
than either of us expected. Let me answer it properly.

On provenance: the table was not copied from any external glossary, so
there is no third-party material and no permission needed. It was
compiled during this work with the help of an LLM-based assistant,
which drafted the zh_CN column from general knowledge of mainland
terminology rather than from the zh_CN tree itself. I should have said
this in the cover letter from v1; more on that below.

That drafting method is exactly the weakness your question exposed. A
column labelled zh_CN reads as a claim about what the zh_CN documents
actually use, and "what a model believes mainland usage to be" is not
the same thing. So I went through all 115 entries against
Documentation/translations/zh_CN/, counting actual occurrences rather
than just checking presence, and found two wrong:

  - build: the zh_CN column had 建置, which is the Taiwanese term and
    appears 0 times in the zh_CN tree. Corrected to 构建 (246
    occurrences).
  - account: it had 账户, which appears once; the tree consistently
    uses 帐户 (8). Corrected to 帐户.

The rest all appear in the tree. A few are low frequency but genuinely
mainland usage, so I kept them alongside the dominant form (e.g. 缺省 2
next to 默认 328).

Both fixes and a provenance note will be in v3:

  本表的zh_CN欄位為對照參考，其用詞經核對核心文件樹zh_CN譯文中的
  實際用法，非引用外部術語表。

On the broader point: I should be clear about how this series was
produced. An LLM-based assistant drafted the initial localization of
each document and this glossary; I then reviewed every line by hand --
checking each terminology choice as a native speaker, verifying the
"update to commit" baselines against git log, reading the text aloud
to catch translationese, and building the docs. That review is where
the substantive fixes came from, including the ones in v2 and the two
glossary corrections above.

The documents this series translates now ask contributors to declare
this, and I should have done so from v1. v3 will carry Assisted-by:
trailers and a note in the cover letter. If you or the other
maintainers would prefer a different form of disclosure, or would
rather I redo any part of this by hand, please say so.

Chen-Yu


葉宸佑 <chenyou910331@xxxxxxxxx> 於 2026年7月29日週三 上午1:54寫道：
>
> On Tue, Jul 28, 2026 at 09:36:24PM +0800, Alex Shi wrote:
> > > +下表同時列出常見的簡體中文（zh_CN）譯法，方便從zh_CN翻譯轉換或對照時
> > > +使用。
> >
> > This is related with copyright. At last, you need to ask the original
> > author's permission, and give a reference here. I guess...
>
> Thanks for raising this -- it turned out to be a more useful question
> than either of us expected. Let me answer it properly.
>
> On provenance: the table was not copied from any external glossary, so
> there is no third-party material and no permission needed. It was
> compiled during this work with the help of an LLM-based assistant,
> which drafted the zh_CN column from general knowledge of mainland
> terminology rather than from the zh_CN tree itself. I should have said
> this in the cover letter from v1; more on that below.
>
> That drafting method is exactly the weakness your question exposed. A
> column labelled zh_CN reads as a claim about what the zh_CN documents
> actually use, and "what a model believes mainland usage to be" is not
> the same thing. So I went through all 115 entries against
> Documentation/translations/zh_CN/, counting actual occurrences rather
> than just checking presence, and found two wrong:
>
>   - build: the zh_CN column had 建置, which is the Taiwanese term and
>     appears 0 times in the zh_CN tree. Corrected to 构建 (246
>     occurrences).
>   - account: it had 账户, which appears once; the tree consistently
>     uses 帐户 (8). Corrected to 帐户.
>
> The rest all appear in the tree. A few are low frequency but genuinely
> mainland usage, so I kept them alongside the dominant form (e.g. 缺省 2
> next to 默认 328).
>
> Both fixes and a provenance note will be in v3:
>
>   本表的zh_CN欄位為對照參考，其用詞經核對核心文件樹zh_CN譯文中的
>   實際用法，非引用外部術語表。
>
> On the broader point: I should be clear about how this series was
> produced. An LLM-based assistant drafted the initial localization of
> each document and this glossary; I then reviewed every line by hand --
> checking each terminology choice as a native speaker, verifying the
> "update to commit" baselines against git log, reading the text aloud
> to catch translationese, and building the docs. That review is where
> the substantive fixes came from, including the ones in v2 and the two
> glossary corrections above.
>
> The documents this series translates now ask contributors to declare
> this, and I should have done so from v1. v3 will carry Assisted-by:
> trailers and a note in the cover letter. If you or the other
> maintainers would prefer a different form of disclosure, or would
> rather I redo any part of this by hand, please say so.
>
> Chen-Yu
>
> Alex Shi <seakeel@xxxxxxxxx> 於 2026年7月28日週二 下午9:36寫道：
>>
>>
>> > +術語表
>> > +======
>> > +
>> > +:作者: 葉宸佑 Chen-Yu Yeh<chenyou910331@xxxxxxxxx>
>> > +
>> > +本術語表列出Linux核心繁體中文（zh_TW）翻譯所使用的術語。zh_TW翻譯以臺灣
>> > +慣用的資訊術語為準。新增或更新翻譯時，請依照本表選擇用詞，以維持整個
>> > +zh_TW文件樹的一致性；若遇到本表未收錄的術語，歡迎提交補丁補充。
>> > +
>> > +下表同時列出常見的簡體中文（zh_CN）譯法，方便從zh_CN翻譯轉換或對照時
>> > +使用。
>> > +
>> > +一般術語
>> > +--------
>>
>> This is related with copyright. At last, you need to ask the original
>> author's permission, and give a reference here. I guess...
>> > +
>>