User:Want/Splitting translatable paragraphs or not?

Since May, User:Amire80 repeatedly ask User:Shirayuki, why divides paragraphs into multiple TUs: First Topic:Y65atzx2j35zdih4 (6.5.2024) and renew Topic:Y98jv594ccbyo676 (27.7.2024)

I stepped into theme, because before I became a translation administrator on MediaWiki.org in 2019, I wrote to User:Shirayuki for clarify what he does and how he does it. And why does he do that.

I did from 2008 year on MediaWiki.org 17298 editations, User:Amire80 15873 (from 2007), but User:Shirayuki, which started 2012 – 343767(!) That's really enough to figure out the best practices how to doing.

I have before 2019 more experiences with Extension:Translate, because I started in 2013 own a multilingual wiki, but I still wrote to him. The result of this is the page Project:Translation/cs, where I summarize all needed informations for translators in the Czech language and recommend pages to priority traslation.

You can see for yourself, by Special:ActiveLanguages/cs that the content in the Czech language is a second position to English original content – only 6 peoples are the base community of translators this wiki into the Czech language.

With all due respect to work, User:Amire80 which is a contractor the Wikimedia Foundation's Language Engineering team (User:Aaharoni-WMF), don't understand how to work with the Extension:Translate.

It's sad. For it I wrote this essay.

Content MediaWiki.org is also created by developers what have not non-native English language, but they code need to be documented. No one will write the documentation for them. They often use automatic translation. From a linguistic point of view, their documentation is not much, but they hope to find a volunteer to do a language proofreading of their text.

Ideally state to be marked for translation is a completed page, but a documentation page is never finished. The text therefore mixes the corrected TU and the text that is just waiting for someone to correct it and mark it for translation.

What is wrong with a paragraph that contains several sentences and is only in one TU?

When marking the text for translation, individual TUs are continuously and incrementally numbered.

If you edit an entire paragraph, the entire paragraph will be marked as invalid, even though you only correct one typographical error.

But the translator does not know whether to make a new translation or not. He has to look at the differences and decide based on that whether to leave the translation as it is or whether it requires a change. And if the paragraph is too long or he doesn't understand it well, he prefers to leave it to someone else.

There are many pages on this wiki, their TUs have been waiting for an update for several years! Translating individual sentences is much easier. They are often repeated.

Often it is enough to translate only those, and there will be one, two or three TUs at the most, with content that requires the translator to have knowledge of the issue.

Well, then a language purist comes along and confuses the original text. A translator who sees many changes in the text is confused by this and prefers to leave it to someone else. The braver one translates it in his own way and is then criticized that the result of his translation is incomprehensible.

If each TU is only one logical sentence, it is much easier to redo the entire paragraph and reformat the page. The editor does not interfere with the content of TUs that are correct, only changes their order, and at most creates a few new ones if the situation requires it.

Language subpages

Do you know what the language subpage looks before Extension:Translate replace the TU by translation version?

Today, the Translate extension doesn't allow 'edit' action for language subpages anymore - now do it redirect to the translator interface. But In the early days, when I start uses the Translate extension, for language subpages was action 'edit' availabled and you were can displayed code of it. It's security potection which won't let you break your wiki.

If view the language subpage code, you don't see the code it's working with - only its final interpretation.

But the translate administrator, before marks the code page for translation, see changes before the original code of the translated page has undergone changes. And that's even if the content of TU doesn't change!

I was thinking about what page I could show it on. I ended up using an older wiki page where I have this extension available, but I don't use it as much as Thewoodcraft.org

I chosed for demonstration of it the IT contacts page.

This page was created on 11/26/2010, long before the Extension:Translate was used. The first marking of this page for translation was done 4/6/2019 – before I was use the old style of MediaWiki translation (the language subpages was standalone pages).

I haven't touched it since then, so this page had to be converted to the Extension:Translate new syntax first. The only visible change we can see is in the syntax of the 'tvar' element. Now, after conversion, the changes are displaying before the translation marking.

Because viewing history and code of the pages is a privilege only of logged users on my wiki, I must to refer you to the image: https://wiki.control.fel.cvut.cz/pub/translate-diff.png

On the left is the change before the conversion and on the right is the change before the new marking of the converted page. The content of TU has not changed. Only code of the origin page.

It is simple. The page behaves like a template. The Extension:Translate goes through code the TU step by step, and if it finds a subpage for the TU in the 'Translations' namespace named by the code of the selected language, it uses it. If not, it will display the original TU content.

Changes to the origin page are included in the translated subversion when the page is marked for translation. Until this happens, the last marking subversion uses if browser call the language subpage. However, the changes after marking of the translationable page, may not be reflected immediately. The request is waiting in the jobs queue, and for its realization it is sometimes necessary to update the content of a TU.

For a wiki that doesn't have a lot of visitors, these changes can be enforced on the server side with a script:

extensions/Translate/scripts/refresh-translatable-pages.php --jobqueue