which translations to maintain?

[slightlyskepticalpotat]

From a quick look through src/qt/locale, many languages we currently have "support" for are 50-90% untranslated. I have several general questions:

• What is the maintenance burden of keeping untranslated languages around?
• Is it worth keeping "translations" around if most strings are untranslated?
• Do we want translators to continue using the Qt Linguist/editing XML workflow?
• Are there any countries where a) english isn't prevalent b) there are many dogecoin users c) the translation is very incomplete?
  (I have been checking using https://what-is-dogecoin.com/nodes/ but don't know how accurate it is)

[daank-c]

Is it worth keeping "translations" around if most strings are untranslated?

I think so. When new Core users come online, they'll notice their language isn't fully supported and potentially come to help. This maybe isn't ideal, but it might do better at encouraging participation than not having the language supported at all.

Do we want translators to continue using the Qt Linguist/editing XML workflow?

I think it works okay but it's not super accessible, and it's my belief that accessibility is at the root of participation. On that note, I was thinking about this recently and decided to reach out to Transifex to see what it would cost to use their service. This is the response I was given:

Thanks for taking an interest in Transifex!

Regarding your open-source inquiry, to 
create an open-source project, the project
should not have funding, revenue, or 
commercialization model and you will be 
asked to provide the following:

- A link to your project's code repository.
- The repository must contain a LICENSE 
  file stating your project's license type.

Here is more information on our open-source
projects:
https://help.transifex.com/en/articles/6236788-open-source-projects.

[daank-c]

I don't mind going through and removing all of the erroneous "unfinished" tags that have been left behind over the years. This would give me something to do, and help us get a more accurate reading on completeness.

[patricklodder]

"unfinished" tags aren't erroneous, so please don't delete those. What you probably mean is "vanished". We keep those on purpose, for situations where just a few words changed (most of 'em) and there is a reference in there what the previous translation was.

[daank-c]

Oh okay, my mistake. I figured the translations with "unfinished" tags, that actually have translations, were left behind unintentionally.

For example from bitcoin_af.ts:

Line 957-960

    <message>
        <source>Rescanning...</source>
        <translation type="unfinished">Word herskandeer...</translation>
    </message>

[patricklodder]

2 things are possible:

1. The translator left it as unfinished
2. Qt lupdate did a fuzzy match on the text - eg when it moved from one module to another - and it needs to be confirmed.

It just means that it needs to be edited by someone that speaks the language. I think a golden rule to translations is: let's not edit anything for which we don't speak the language, unless it's scripted by Qt Translations tools.

[daank-c]

Noted. Thank you for the clarification.

[patricklodder]

What is the maintenance burden of keeping untranslated languages around?

None, because updates are scripted. The reason for removing untranslated dialects is because it is confusing.

Is it worth keeping "translations" around if most strings are untranslated?

I'm neutral on this.

Do we want translators to continue using the Qt Linguist/editing XML workflow?

I don't mind moving to transifex if translators believe that that gives a better experience.

Are there any countries where a) english isn't prevalent b) there are many dogecoin users c) the translation is very incomplete?

I feel that we cannot include (b) here, because maybe there aren't many users because there is no complete translation. So I think that it's better to make decisions based on how large the total population is we can reach with a translation.

[slightlyskepticalpotat]

You raise a good point about (b). From a quick check, I think some countries are

• India (Hindi, almost completely untranslated)
• Pakistan (Urdu, completely untranslated)
• Bangladesh (Bengali, no file)
• Ethiopia (Amharic, no file)
• Egypt (Arabic, PR at Translate arbic #3439)

About what I expected, countries with sizable populations that interact less with the english-speaking internet.

Large countries that already have a mostly (>80%) finished translation:

• China (Chinese)
• United States, Nigeria, Philippines (English)
• Indonesia (Indonesian)
• Brazil (Portuguese, see prior discussion)
• Russia (Russian)
• Mexico (Spanish)
• Japan (Japanese)
• DR Congo (French)

[slightlyskepticalpotat]

Out of the 10 most spoken languages in the world (arbitrary bar), we are missing Hindi (3), Arabic (6), Bengali (7), and Urdu (10).