Hacker News Reader

Comment by vaylian | Hacker News Reader

I really wish for a language setting that says: If the original text is in one of the following languages, use the original text. I might not be equally fluent in all of those languages, but reading the original text is better than reading a translation most of the time.

View on HN

You’re in luck!

https://developer.mozilla.org/en-US/docs/Web/HTTP/Reference/...

I wrote about it once; https://dijit.svbtle.com/trusting-the-user-they-know-what-la...

Outlook web actually respects this, which is refreshing.

Unfortunately, that just specifies which languages you prefer, regardless of what the original language is.

I speak Dutch and English, but have English set as my preferred language, because most of the time that multiple languages are available, that is the original language. However, sometimes I'll be visiting a Dutch site that has a (usually badly-)translated English version available, and I'd rather get the Dutch version.

I think thats bad logic on the server side and not to do with the concept of Accept-Language though.

I would think the ideal prioritisation should probably be:

* Exact (en_GB)

* Exact (fr_CA)

* Soft (en_*) // Any variant of English you might have available

* Soft (fr_*) // Any variant of French you might have available

* Translated. (to first preferred)

for Accept-Language: “en_GB,fr_CA”

I don't think Accept-Language supports "Translated", "Exact" or "Soft"? But yes, that would match the language setting that GGP asked for.

This is expressed as the quality parameter. To make the system GP proposed work atop HTTP content-negotiation, a server operator should assign a very small number to the automatically translated content.

Isn't that parameter sent in the Request?

(Also note that I wasn't talking about automatically translated content necessarily. Manually translated content still often is of worse quality than the original language.)

It is; but also there is a server counter-part with its own quality weights. If that does not make sense, play with the code in http://p3rl.org/HTTP::Negotiate and read the HTTP spec.

TIL. Well, then I wish that was widely implemented :)

Isn't that what `Accept-Language: *` is for?

You could probably also trick websites into serving the original language by setting a language they probably don't support (how many websites have a version available in Akkadian anyway?) but you'd need to pick something that the Google Translate widget found on many websites doesn't pick up and try to use.

No, because then it will probably serve it in the original language regardless of the languages I speak. If the original language is French, I'd still prefer to have it in English.

(To be clear, with how widespread English is, and how often Dutch sites are Dutch-only, this is a minor issue - having English, then Dutch as my preference is usually good enough.)

Modern browsers with auto-translation features always offer the option to "Never translate X" — never translate a language the user speaks.

If websites really need translation (which they very likely do not!), this should be the norm as well.

MacOS has this feature under Settings -> General -> Language & Region -> Preferred Languages. Here you can have an ordered list of languages you wish apps and websites (I _think_ only works for Safari) to use. I don't know how widespread this is in other OS'es or browsers.

That's just the standard Accept-Language header, not unique to Safari or macOS at all. It's up to the server to interpret and respond to the header, and they mostly don't do what gp wants, i.e. they usually serve the translated language with a higher q rather than the original language even if it's in the list (and it's mostly ignored anyway in favor of geoip/manual region/language picker).

It is not the accept-languages header - it affects the accept-languages header. Like the parent said, it is also used for apps (which includes your browser's UI).

This has been OS-independent since the dawn of time (see the Accept-Language header specification), but sites rarely use it

I think the date might be later definitely after the adoption of unicode characters.

Reply on news.ycombinator.com