Adopt Translation & Localization Management Platform for docs

[MaxymVlasov]

This is a Feature Request

Support of Translation & Localization Management Platform (TLMP) for docs

To effectively make and maintain translations, we need to adopt tools that were created for it.
There a bunch of solutions, like Transifex, Crowdin, and others.

After a quick research at the start of 2023, we chose Transifex for PoC as a free and reliable solution, but in the end, it's no matter what tool sig-docs chooses, it will be much better than the workflow that we have now via git&Github.

What would you like to be added

There are 3 ways, from better to worse:

1. Full integration. Integrate any TLMP with k/websites and move localization work fully to the chosen TLMP, including docs reviews and CLA sign. Add TLMP Github integration and auto-merge changes sent by TLMP integration user (provided by TLMP or you can setup your own)
2. Particial integration. Translate and review in TLMP, but users still need to sign CLA in Github
   1. Fix the EasyCLA co-author issue, mentioned in Linux Foundation issue, test PR
   2. Create a script that will be take by API's last editors of each line for the changed file and add them as co-authors in git
   3. Auto-merge such changes if EasyCLA check passes, (review already done in TLMP)
3. No integration. Same as p.2, but 2.3. replaced with:
   3. Manually approve and merge such PRs in Github, if EasyCLA check passes.

Why is this needed

When you write docs or code in 1 language, you work only with the current state of docs - everything that you need to track is tracked by git.
When you translate something - you deal with 2 sources of truth: original and existent translation, which can be partial, outdated or placed in locations that have already been moved/removed from the original. + other edge cases. There is no easy way in git to check what changes in the original also should be revisited in translation, so mostly every translated doc became unsupported from the moment when it was merged - as sometimes simpler to redo a translation from scratch than to figure out what changes are needed.

Long story short: git is just the wrong tool for translations. It is as bad for this as using .zips for VCS or trying to send a letter by pigeon mail to another continent and hoping for 3 workday answer.

Also, tech-writers, students, or newbies which'd like to contribute, in most cases have no or little knowledge how git and GitHub works, they don't have a GitHub account, and so on. Those are just not intuitive tools for non-techie folks.

So, what good enough TLMP will provide:

• Tracking changes in the original and asking for new translations for these changes
• Provide suggested translation by Google Translate/DeepL/etc by 1 click, which should be reviewed at least twice: by a user who added that auto-translation and by the reviewer(s)
• Have built-in CLA support
• Have a Github integration
• Have an easy to understanding for translators interface
• Provide easy-to-find not translated, and not reviewed strings/files
• Able to provide/keep a few translation suggestions for a single string for the reviewer choose
• Easy to signup and sign CLA from the start mechanism
• Conform with all LF and CNCF regulations (or change those regulations to conform with TLMP)
• ... ?

Comments
@sftim asked to add it as an issue here, to be able to track work on it.

Related Linux Foundation issue

Screenshots of the LF issue in case you can't see it

Read msgs from bottom to top

P.S. That started as a Ukrainian localization team initiative, but we were blocked from the legal perspective of merging back that kind of change from TLMP.

[sftim]

/area localization
/priority important-longterm
/triage accepted
/kind feature

[sftim]

For Transifex: if Transifex commits a change, is there any license or copyright asserted by Transifex?

(if not: I think we could make a tool that allows contributors to adopt those changes and confirm, through the tool, that their CLA applies to the contribution(s) in that commit).

[Andygol]

Consider integrating Crowdin (https://support.crowdin.com/enterprise/authentication-settings/) along with Transifex for translation purposes. Crowdin offers a comprehensive set of authentication settings, as illustrated in the image below:

Additionally, Crowdin provides a versatile range of translation tools, enhancing the efficiency and engagement of contributors in the translation process. The image below showcases the various tools available:

Encouraging the use of Computer Assisted Translation (CAT) tools like Transifex and Crowdin can significantly streamline the translation process and boost overall contributor engagement. It would be beneficial for CNCF to explore and embrace these tools for a more efficient and collaborative translation experience.

[windsonsea]

I like it. Translation memory is always asset!

How about to introduce chatGPT that can provide better results than traditional machine translations.

[sftim]

Let's have one issue per change we need to make; if we'd like to have a wider discussion, see:

• Generative AI for Kubernetes docs
• Translation tooling for Kubernetes localization

as repo discussions

[sftim]

@MaxymVlasov & colleagues: do you have a reply for #45175 (comment) ?

[OleksaBaida]

Hello @sftim .

if Transifex commits a change, is there any license or copyright asserted by Transifex?

According to https://www.transifex.com/legal/terms/ the answer for quoted question is

Transifex does not assert any ownership rights over the content a person submit, including text published via git repositories for localization. The terms specify that while using Transifex services, you retain full ownership of your content ("your words"), and Transifex only requires limited rights to perform the services requested by localizator\customer. This includes actions like hosting, sharing, or otherwise processing your content as directed by you, without claiming any copyright or license over it​

сс @MaxymVlasov @Andygol

[OleksaBaida]

@sftim pretty similar conclusion for CrowdIn mentioned by @Andygol https://support.crowdin.com/terms/

Clients are responsible for ensuring they have the necessary rights to their content. Crowdin asserts no ownership over client data, which includes text submitted for translation or any other purpose

cc @MaxymVlasov