Hacker News Reader

Comment by WalterBright | Hacker News Reader

In the 80s I worked pretty hard to support multiple languages with the Zortech C/C++ compiler. Error messages were switchable between English, German, French and Japanese. Translated versions of the manual were made.

The trouble was, I had to hire translators. They weren't programmers, so the translations were (so I've been told) peculiar. When I'd modify the compiler, trying to keep the translated text in sync was a nightmare.

The last straw was when I found out that essentially none of our customers used the messages in their native language. They preferred the English versions.

So I just gave up on that with D. (Although the D language itself has excellent Unicode support, the user interface is all in English.)

Some members of the D community have taken the initiative to create documentation in their native languages, which is great.

View on HN

Interesting story! Thanks for sharing.

I'm consistently impressed by Microsoft's approach to translation - their online database of annotated translations - is an invaluable source for translating and just talking about computer/software stuff in non-english languages. ( https://www.microsoft.com/en-us/language they made their entire translation database available online).

I do some work on a compiler and as much as I'd like to enable people to use it to teach young schoolkids in their native language (with localizable keywords + error messages), the implementation/maintenance burden would be massive and then people wouldn't be able to share source code so easily globally.

It's hard to balance - I like languages, and usually my code isn't in English, but...yeah, I can only go so far in practice.

While Microsoft still is the best in this regard, they too are not perfect. I routinely see very strange thing when working on devices that aren't in English. Things like translating the word "manual" in the camera app (as in "manual settings") into the equivalent of "manual" as in "instruction manual".

Strangely enough, open source projects tend to actually have better translations, since they are more often than not done by developers and users of the software, who have far more context than the random translator with the lowest hourly rate that most companies go for.

> open source projects tend to actually have better translations, since they are more often than not done by developers and users of the software, who have far more context than the random translator with the lowest hourly rate that most companies go for.

Thanks for sharing this. I totally agree with you.

I've had the idea of making a fully localizable programming language for quite a while too, but like you, I've never actually done it due to the amount of maintenance a project like that would require.

My idea was to make it possible to transpile between all the language variants and allow machine translation of variable names and comments. That way, after a new language variant were implemented, users (primarily teachers) could take any compatible code in any language they even remotely understood and translate it, then do just a bit of cleanup and be able to contribute it back and use it when teaching.

Then I'd build a repository of these small programs, each with a set teaching goal in mind, translated into as many language variants as possible.

But now this is starting to sound like months and months of work, and since I can't focus on a single side-project for even a few weeks, is unfortunately never getting done...

As a french person, the Microsoft website in french is absolutely terrible, I just always go for the "English" button

Ok, it's useful to know that they're not equally good in all languages - I had incorrectly generalised!

It’s much worse now - anyone who uses any program that supports localized error messages should know by now to switch to English to have a hope of googling the problem.

I suppose unique error codes could help here.

Boy, am I glad for your comment. I hadn't even thought of this, but now, I think I will use unique error codes for this reason.

> unique error codes

Today those are not necessary. With how good search is these days, what works great are crafting the message strings so they are search friendly.

Together with bad translations it means that there are error messages that make zero sense and you can't find help online. Indeed, the most important localization feature is turning this off.

Thanks a lot for sharing your experience it's really great to learn more about FLOSS and its community.

There were some fun moments in this. In attempting to translate "destructor" to Japanese, we'd get "death tractor". Now, personally I felt that "deathTractor" was a far more apropos term than "destructor" (sorry Bjarne). For years my circle of colleagues called them deathTractors.

Reply on news.ycombinator.com