Hacker News Reader

Comment by atemerev | Hacker News Reader

My take on internationalization is that I want to see less of it.

I am Russian living in Switzerland, where people speak French, (Swiss) German, Italian and Romansh. I often travel to Barcelona, where people speak Catalan and Spanish. I don’t want any of these languages, I want all sites in English.

This is hard to achieve, as nearly everybody uses geolinked internationalization. It is awful. Especially here in Switzerland — almost every site tries to show me the German version by default (most of Swiss cantons are German-speaking, but mine speaks French).

Please just use English. Or if you must, enable language selection in users profile, do not try to detect it from geographic location.

View on HN

My pet peeve is when I have an application crash and the error message has been translated. Now googling it is extra fun, not to mention bug reporting.

Fortunatly with Linux, I usually can get the proper message with "LC_ALL=en_US.UTF8" if I can, and desire, repeat the exact same error.

Now I understand that this is done for openess and diversity. But technical terms should not be translated. I'm french, and I accept that english has won as the business language, and the IT language. Let's move on.

Honestly, we are wasting humanity potential for not having a common language. I really, really wish we would all learn english as the second language, and that all laws, forms, labels and signs would be translated to english everywhere (as an addition to the local lang, not a replacement). It's an imperfect language, but it won half of the world, and the other half will have an easier time learning it that us learning theirs. Plus it has a very simple alphabet, and is already used for science, business and international politics.

At this stage, I would gladly agree for all the french books and movies being burned and lost forever if it means one language, any language really, becomes magically the main one for humanity. We have too many communication problems.

But since magic doesn't exist, what about getting off the ego train and just adopt the imperfect status quo and make something awesome with it for a change ?

English in terms of number of speakers has already reached the point where the number of ESL speakers outstrips the number of native ones by a good margin. The simple alphabet is a plus. I wish the spelling wasn't so godawful.

Yeah but I don't know any language used by hundred of millions of people that doesn't have many gotchas.

Could be worse. The future language could become a mix of emoji and markup language :) ;) :D

> The future language could become a mix of emoji and markup language :) ;) :D

You're saying this like it isn't already happening.

:: insert flail/panic emoji here ::

😱

To fair English has big pieces of French and Latin vocabulary.

One of my favorite aspects of the English language is that not only is it shameless about stealing other languages' words, but it seems to take it as a point of pride. It's a good feature to have if you want something global- it means that useful words will accumulate without regard to origin, which benefits everyone.

"We don't just borrow words; on occasion, English has pursued other languages down alleyways to beat them unconscious and rifle their pockets for new vocabulary."

https://en.wikiquote.org/wiki/James_Nicoll

Noted reviewer of science fiction and fantasy.

and also by extension, Greek (directly in many cases -- almost every english word that starts with 'ph' was a greek word -- but also a large portion of medical and mathematical terminology)

That's because in latin 'ph' was used in words from greek, but spanish uses 'f' AFAIK.

Fun fact: russian language had no 'f' sound prior to cyrillic alphabet which has a letter for it.

I suspect english words starting with 'ph' come from french, french words with 'ph' used 'ph' instead of 'f' (same sound) to emphase the fact the word borrow greek root (but who care the word come from greek, does it make it easier to learn? one more shitty rule of french).

Granted I don't have hard data on the "ph" fact, the source is pretty credible in my opinion. The fact appeared in an article written for the British Council, co-authored by Martha Peraki, a Linguistics PhD recipient from American University.

Heuristically, the "rule" passes the eyeball test: philosophy, physical, photo, phrase, philanthropy, phobia, phage, phalange, phalanx, phallic, phase, pharmacy, phantom, phenomenon, phone, photons, photosynthesis, physician, physique, phytoplankton, so on and so forth

The overwhelming majority of these words are greek transliterations https://www.thefreedictionary.com/words-that-start-with-ph

(in fact I am having a hard time finding even one word that isn't -- i think "phreak" is the only one I can find)

Geographic location is incredibly ineffective anyway depending on your ISP and other factors.

Rather than defaulting to English, why not default to whatever the browser's OS is set to? At least for JS apps, there's navigator.languages, and there's an Accept-Language header for both language and locale for standard server-rendered plain ol' websites.

Geographic location really does seem like the worst possible option to use.

Yet Google, Microsoft and Apple use it. :( At least with Google, you can usually log in to your profile and set it to English; with Microsoft and Apple, you are out of luck.

That's one big problem on the web, IMO. Google in particular does a lot of things that are questionable UX-wisely, to put it charitably, and then a lot of web shops try to copy those things without giving them any critical thought.

Is it really the worst possible option? It’s often a pretty good guess, and accept-language is similarly unreliably set and can be fiddled with by intermediaries. Navigator.language doesn’t help you until you’ve already served and run JavaScript.

The right solution I think is to make a reasonable guess and then allow the user to control it. None of these guesses are going to work for everybody on their own.

I'm unaware of any intermediaries that fiddle with accept-language, so I can't really speak to how reliable it is. I can say that due to the ISP I use geolocation typically gets my location wrong by somewhere between 650 - 1600 km (400 and 1000 miles).

Defaulting to OS language can be terrible for organizations where the working language is different than a local language. Lots of programs do it and they always default to the language the OS was installed in, not the one that is currently used by the user.

I work in an international organization in Czech. The local IT team installs Windows in the local language, then sets it to English (which is fine, mostly). Half of our employees don't speak Czech well and changing weird tools to English is a constant and difficult hassle for those who haven't learned the basic words yet (printer drivers and small open source tools are the worst).

Browser language can be difficult as well. Half of the population is bi- or multilingual.. choosing which language to use for a browser (for me: German or English) affects how well Google Translate works - which I use daily. (Somehow, big American companies don't get the idea that people can live in multiple countries at the same time or speak multiple languages).

> I work in an international organization in Czech. The local IT team installs Windows in the local language, then sets it to English

That doesn't make sense. If the OS is "set to English", how do programs know to run in Czech?

Anyway, we just install Windows in English. (British English, so the dates appear the right way around and there's a chance of metric measurements.)

User / Domain profiles that set the language of wherever you login. The "original" installation remains in Czech.

Yes, it would be preferable to have the installations in English initially. But we weren't always international and not all of our employees speak English well (this is common for older people - former UDSSR ..., admin staff and people in roles that don't require a college education). There are still legacy tools in Czech, fully Czech teams (who obviously work partially in their native language), a need to work in the local language (procurement, legal topics,..) etc. It's just not economic to completely enforce English.

just FYI Czechia was never part of USSR and it's quite offensive to tell it to Czech person after being occupied for decades by Soviet military. that's like saying Afghanistan and Iraq are US states to people living there, just because US occupied their country

You are correct, I was sloppy writing that and apologize.

Growing up, I learned that Czechoslovakia was part of the "Eastern Bloc" and we were never taught much about the individual countries and their peoples. Moving there taught me a lot more about the country and its neighbors' history and culture, and also my own culture.

Although I'm not sure about "Central Europe".

> Although I'm not sure about "Central Europe".

This term predates the Iron Curtain.

it all comes down where you think Vienna it's, because it's for sure more eastern city than Prague, but i don't see people calling Austria eastern European country, which is political from before 1989

i have no problem ignoring central Europe when people will be consistent about geography especially regarding Austria, but since nobody it's putting Austria in eastern Europe then we have to live with Central Europe, after all Europe ends at Ural, so calling Czechia eastern Europe it's quite a stretch

people can live in multiple countries at the same time

As a silly American, can you explain how you live in multiple countries at the same time? Do you mean you, for example, live in one country and work in another?

I live in Spain and Switzerland. I spend most of the summer in Spain, and the rest in Switzerland. I work 100% remotely, so I can live and work anywhere (not seamlessly, as I need visas and residency permits and everything else, but still doable).

My wife is a scientist, so we travel a lot and change places of living. So far, I lived in Russia, Switzerland, Italy, Spain and Denmark, and about 2 months in the US.

Working at CERN is one example

you can live in Bratislava which it's less than hour drive from Vienna where you can work every workday, plenty of people in Europe live nearby border and spend lot of the time on both sides speaking completely different languages

This is a favorite topic of mine. There are several multi-country metropolitan areas in Europe, for example Malmo-Copenhagen and Görlitz/Zgorzelec. Many cities have smaller suburb cities across the border, like Strasbourg-Kehl and Geneva-Annemasse. And the Benelux countries take this to the extreme, of course.

Even Berlin, which is pretty far from the Polish border, have commuter trains to Poland. That's right, the Berlin metro ticket will take you to Poland.

As a young adult, I studied in a country (had a flatshare there) and took the train (2hrs) at the weekend and during the holidays to stay with my parents. I guess lots of people close to borders know family situations like that (parents, partners).

Now, I have a job with 90% business travel between two countries (but same cities). Have small apartments in both. It is just more economical and comfortable that hotel room living.

There are also often reasons not to completely move to a country, even if it's close (different tax situation, having to switch social security systems which affects your retirement, wanting to stay inside a certain social security systems which requires a place of residence..)

Changing browser language settings is a couple clicks. Are your systems not user configurable or something?

Rather than defaulting to English, why not default to whatever the browser's OS is set to?

As someone who develops multi-lingual web sites, I can tell you this doesn't work.

It should work, but it doesn't. We test with actual users in their home environments, and... people are messy.

- The vast majority of our users accept the default language (English) when setting up their computers, even if they don't speak English.

- In households where the parents are immigrants, very often the children set up the computer, so it's set to English, which is the child's preference, even if the parents don't speak English.

- Poor households often share a phone or a computer, and each person may not speak or prefer the default language that was chosen by the person who set up the computer.

In the idealized world that Silicon Valley imagines where 1 person uses 1 computer with 1 owner and speaks 1 language, it's fine. But in the real world, it's just a mess.

Ah that makes sense. I'll stop complaining about this now.

I'm not saying abandon multilingual capabilities, I'm saying Rather than defaulting to English, why not default to whatever the browser's OS is set to?.

This was in response to a comment chain suggesting that geofencing default language is a bad option, because of how inaccurate it can be.

> - In households where the parents are immigrants, very often the children set up the computer, so it's set to English, which is the child's preference, even if the parents don't speak English.

I fail to see how any default language setting will help in this scenario, assuming they've immigrated to an English-speaking geographic region.

> - The vast majority of our users accept the default language (English) when setting up their computers, even if they don't speak English.

I suppose this is one scenario where geofencing might work better than OS default language, but then again, based on my experience it's just as likely to be wrong. At least with English it'd be an interface they're accustomed to seeing, and would have better luck finding the website's language selector, rather than something more arbitrary :/

This sounds like something that web browser UX experts could help to fix.

They already did. Every major browser allows you to configure the language you speak.

Of course it doesn't do much because sites do very stupid things like use geolocation.

Firefox (and I suspect other browsers) allow per-profile configuration of the language, independent of the OS.

In the scenarios you describe, it's hard to see how geolocation makes things better, as it has all the same problems, except for the "installed in English because didn't change the default" where you now have a chance (but no guarantee) you'll select the correct language.

And unlike configuring the language in the browser, there's no hope geolocation could even possibly work correctly.

>In the idealized world that Silicon Valley imagines where 1 person uses 1 computer with 1 owner and speaks 1 language, it's fine. But in the real world, it's just a mess.

In Silicon Valley geolocation works fine, but it's a terrible, terrible, TERRIBLE idea everywhere else.

If you do do this then please don't use keyboard language as the language. Use ui/display language instead. Looking at you, Microsoft.

This is fromt he point of view of someone that is fluent in English.

I assure you that in countries like France, Italy, Spain, etc... this is true but for a small percentage of the population.

It makes no business sense for a company not to localize a website or a service, if your market has a substantial number of non-English speakers.

But yes, for my own convenience (unironically), I wish that every website had a base English version that you could access with a single click.

For example I was looking at some Microsoft webpage with troubleshooting or API information (can't remember). By default it was shown traslated based on my location, but there was a prominent "Read this in English" switch at the top. Top marks.

The problem is not the localization. It's choosing the default language based on anything other than explicit info provided by user-agent - that is, URL and Accept-Language, in that order.

French here.

Yes, the english level is terrible in my country. Unfortuntly, translation is imperative if you want to enter the market. Which sucks, because it's a huge waste. There is not benefit to it: a UI is not like a book, you don't lose some quality in translation. In fact, most UI are worse once translated because english is a short and to the point language, which is what you want for buttons and menus.

How many people in France have their computer / browser set to English, but don't understand English?

I could see that being more common in the early 2000s, when it was probably easier to find a pirate version of Windows in English, but now...?

I can't speak about France, but my experience in the U.S. is that recent immigrants who can't speak English still set their computers to English.

For some it's because they're so used to the non-English versions of web pages being so bad or abbreviated that having an English page they can half read is often better than the truncated "good enough" version provided in their native tongue.

But in that case they would want to have the English version of the website anyway.

At least get it right. There isn’t much point in serving a German page by default in Geneva.

I never understood why sites try to do geo based localization. Every browser in capable of setting the requested language in the request headers so why not use that ? Technically much simpler as well. And the user can just change the browser preferences and all sites would magically switch to the right language. Hey I can dream right ?

A number of reasons I work for major conglomerate which has over 200 brands and operates in getting on for 200 countries.

In the USA for one brand they sell different products to the UK than the USA for another brand they use a different brand entirely in the USA to the rest of the world.

There are also legal reasons why you might need to geo target.

That displays different content, but the Accept-Language header (with locale specialization) isn't about the content, its about the language the content is displayed in.

That’s not localization, though. That’s market segmentation based on location detection, which is a different thing, strictly speaking.

I mean, they could still show me the UK-only site if they determine that I am in the UK, but show me the UK site in German, with German units, dates and time codes, etc., because that’s what my browser might tell them that I prefer.

Have you any idea how much work that is and how many things can go wrong I might have a dozen or more locales / languages and if you then want to see any locale in any language that's a huge amount of different urls

Ok so you want German content in the UK do I show you the German locale and products legally available in Germany? the UK locale and products legally available in the UK translated into German?

And do I age gate you based on German law or English what happens if I show you a BOGOF offer do I obey German law or English What happens if google messes up. What happens if your in Quebec and I accidently show you an illegal competition What T&C's do I show you

No need to have different urls for different languages. If you detect that the user's location is in the UK display the UK site, comply with UK laws etc... If that same user sends a German Accept-Language header, show them the UI in German. Doesn't seem much more complex that what you're already doing...

You might want to translate the url for any number of reasons mostly SEO related.

But the main problem what if google gets confused and messes up and lands you on the wrong potentially problematic url - do you want to be the developer explain to the General counsel why the company is in trouble.

You can't go wrong with ?p=564. You might want to skip 666, 13, and some other numbers too, though. :D

Yes, those are the worst.

Sometimes I can't buy books or movies or games in the language I prefer (English), I have to wait for (usually poor) localization. The number of times I have saw the dreaded "this product is available only for the residents of US and Canada", I can't even understand, why my money is not good enough for that.

> why my money is not good enough for that.

It's likely not that they don't want your money, but rather that they don't have the legal right to sell you a piece of content in exchange for your money.

If I am visiting another country, I can buy a physical game or movie in a brick and mortar store. There isn’t any reason it shouldn’t work for online purchases.

If you use your laptop in another country, you'll also be able to purchase geo-locked content with your foreign IP address. (Unless they reject foreign credit cards, but that's more to prevent VPN's, and you're just a false positive.)

The way copyright gets licensed per-country is pretty dumb in the age of the internet, but it's not correct to blame individual retailers for this. They would love to have a larger customer base.

Yet I can go to <country>.amazon.com and buy it and have it shipped here. Is Amazon breaking the law?

Actually, there's a lot of stuff they won't ship to you. I had to use a forwarding service to buy video games from amazon.co.jp for many years, although more recently they've started to allow foreign purchases on most video games. No idea what precipitated the change.

I imagine this is mostly a legal gray area that no one cares enough to litigate. The number of people going to foreign sites to paying international shipping prices and wait weeks for their products to arrive is too small to worry about.

I have now several DVDs where the first menu asks me to select one of 6-8 languages. Then it shows me the unskippable copyright ads (you wouldn't steal a car blah blah blah) localized.

Then we get to the main menu which is English only, regardless of the language you selected. The movie itself is in English, subtitled in at most 3 languages, reachable only from the remote control button (ignoring your first language selection), and sometimes these 3 are different from the 6-8 copyright ad languages.

It's fun to see the copyright ad in Russian or Italian.

Developer cleverness/arrogance. “Users are stupid” mindset. Product designers thinking they know better than their users what language they want.

It’s infuriating to travel overseas and, without changing anything about my computer’s setup, having to deal with web sites overriding my chosen preferred language. Please, just stop doing this.

In my experience these are often configured wrong, because browsers are guessing settings based on the locales used by the operating system, and because users don't know how to configure it because it is really hidden away.

For the cases I see it is technical minded people that speak another language than Danish, that buy (or inherit) a laptop with Windows on Danish, so the browser infers its settings from the operating system, and now you can no longer trust this header.

But it might be just as wrong as guessing language based on IP-location, so there are not really a good out-of-the-box solution, so every website will have to build their own selector maybe with input both from the browser headers and from the IP-geolocation.

Browsers should just make it more obvious and easier to configure. But a lot of websites respect this, I've found. You'd be surprised.

Chrome makes sure to suggest translating a page. Why doesn't it suggest to change my language preferences as well? Rather than just suggesting 'ignore translating for this language' as the alternative.

Yes! When I head to docs.microsoft.com they show me a really bad machine-translated site of the C# API. And it is hard to change to the original English site. I am a software developer; I know English, thank you very much!

One of the few things where Google Search is really awful for me personally. I've set it up to get search results in German (country of residence) and English (everything else). But I haven't found a way to communicate to it that for anything technical, I have a very strong preference for the English results.

If I search for "std::vector", I get https://de.cppreference.com/w/cpp/container/vector as the first result and https://en.cppreference.com/w/cpp/container/vector as the second. The German translation is just awful and completely trips you up if you know English. I don't understand why it keeps pushing them up.

My previous job had me traveling between Germany and China; during that time, Google's ridiculous localization of search results was becoming unbearable. I ended up permanently overriding my default search engine to point to this URL: https://www.google.com/search?hl=en&q=%s.

(It broke auto-suggesting search terms in the omnibar, but the actual search results are again sane.)

Me too. I've had to start using DuckDuckGo, NOT for privacy reasons, but it get around Google's lunatic forced geolocated search.

You cut to the core of the problem; user-preferred language for content is arbitrarily context-dependent.

For most search queries, identifying the user-preferred language is quite easy though: It's the language the user used for his query.

Of course, Google goes to great lengths to ignore this hint.

Playing with curl, it looks like docs.microsoft.com is just using Accept-Language. Is your browser not set to English?

I use English most of the time, but the AWS translations from English to German got really good lately.

This is a good idea in theory, but completely unworkable in practice. Most users do not have the skills or know-how to navigate whatever Accept-Language setting their device is exposing.

But yeah, sites should have a way to sticky it once you have a profile and log in etc., but most users aren't going through that path for most sites.

In Firefox: Preferences -> Language and appearance -> Language

Exactly. That's too complicated for a large portion of browser users. I'm not kidding and I'm not dissing them.

They shouldn't have to fiddle with language settings. That's a tech failure.

Isn't the default value whatever the OS is set to? Your "average user" shouldnt need to fiddle anyway.

What is actually a tech failure, is assuming I understand Spanish if I happen to be in Spain. That sort of ham-handed, short-sighted approach drives me nuts.

You only need "fiddling with language settings" if you want something other than how your operating system is configured.

Google actually did this at one point. I was in the Netherlands when I signed up for Google+. I don't speak Dutch and hadn't changed any of my language settings - the only way they knew was via my IP address. The UI was mostly in English, but all my circles (or whatever they were called) were named in Dutch.

I get the weirdest language bugs with Google: - The info box on the right hand side of the search results intermittently shows up in German - A suggested event (a football match) in the UK (where I am) had both team's names transliterated into Cyrillic - An airport on Maps consistently shows up in either French or Finnish (or both!)

If this is too complicated, it is the fault of the user agent. That's a different issue than whether a website should respect a requested language if it is available. You're arguing that the user agents are flawed, so we should have flawed content delivery to compensate. This is how flaws get folded into systems and magnified over time. Just fix whatever flaws exist in the user agents, language selection based on http headers is a pretty good solution for the half of the problem it covers.

Presumably they wouldn't have to. If they are running en-US Windows, they get en-US websites. If they are running fr-CH Windows, that's the websites they get. You don't really have to worry that someone running ja Windows can't read Japanese.

(Of course, you'd still expose the option to change this from the inherited default in the browser's preferences)

Yes you do. Plenty of people using an OS in whatever language can barely speak a word of that language.

Most commonly they'll use an English OS for whatever reason and want to browse websites in their local language.

This is why you'll find that a lot of website language detection tends to discount English specifically. I.e. if your Accept-Language is "en" in France they'll probably give you French, but if it's e.g. Japanese in France they might think that's a "real" setting and give you Japanese.

Or, as I'm sure a lot of non-English speaking technical users here can attest to, they like to use an English OS UI on their laptop or phone, but speak the local language and want browsing a local website to just work. Non-technical users do that too, but have no idea how to tweak their Accept-Language independently from their OS language.

Really, do armchair commentators here think that programmers at major websites like Google that have enough data to know how people are likely to switch languages from the default haven't thought of just implementing what's literally the most simple thing you can do, which is to obey Accept-Language?

BTW. I accidentally discovered that in case of accessing google.com from non-English speaking country, `en-US` is treated as non-reliable (probably because many devices have this set as default), whereas `en-IE` for example is treated as reliable. Also appending any non-English language after `en-US` does the trick AFAIR.

Basically `en-US`-only Accept-Language seems to be treated as "perhaps default language of a device and user didn't change it, we should not rely on it" and IP-based localization is applied in that case ("in France? ok show French version").

In approaching 100% of cases the operating system pesters you to set a language+locale during setup, and the browser propagates that via Accept-Language. The advanced use case is the user who wishes a set of > 1 weighted language ranges; e.g. de-DE most preferred, then English in any culture, then Latin American Spanish.

Doesn't the browser pick up the OS language setting? That's going to be correct for most users.

And browsers could make the override easily discoverable in preferences (at least on MacOS, app preferences are in a standard place from the user's perspective: File -> Preferences).

Tiny correction, the standard is actually `AppName` → Preferences under macOS X.

That's actually not correct for most users. Most well-known case is India, where the OS language is usually set to 'English' for $REASONS, but people still prefer content in Hindi. (Or Urdu)

Guessing what language a user actually speaks is not straightforward.

What are the reasons?

It still doesn't make sense to use the same behaviour in Europe, where in my experience the OS language is almost always the preferred language of the user. (Especially on a mobile.)

In my experience, most European sites will/should consider 'English' as default/unset/unknown. So when in German,y, if the header says French, give French . If the header says English (US or UK), give German.

What am I supposed to do, as a native English speaker living in Denmark?

I've tried en and en,da, but I almost always get Danish in Denmark and the local language if I travel.

FYI, Irish (en-IE) instead of en-US works fine for me, at least on google_com, to reliably get English content.

Several things, AIUI

1) Inertia. The Indian market is largely English-speaking, so many products came to market in English first, and now people are used to it.

2) Colonialism. Using English still marks you as "educated" and "more respectable". So, to not look like a yokel, everybody sets their UI to English. Content, which is more ephemeral, still is successful in Hindi/Urdu.

3) Localization is hard, and often the English localization is better - but content isn't localized, its first language is the native language.

IIRC, most browsers do default to the OS language. It's just not the correct choice in all cases :)

Websites doing weird things is websites trying to be smart and failing on top of that.

Same happened in Pakistan, the higher up the social ladder you go, the more likely you stumble on somebody who simply can't do written Urdu, or even speak it.

That all makes sense, thanks for clarifying.

How is having a significant number of users be served the wrong language for every site they visit, then having to navigate and create a profile and set the language for each site individually in a different way, in any way an acceptable solution either?

If for historical reasons Accept-Language needs to pass through the OS locale, which is wrong and can't be changed, why not just standardize an optional User-Requested-Language(s) header, that only appears if the user explicitly sets it in the browser?

The current solution is awful and improving it doesn't seem like rocket science but there seems to be zero interest in fixing it. Stuff like this is why closed platforms like Facebook probably have a brighter future than the open web.

I think your comment is kinda funny. Reminds me of California, a lot of effort goes into to providing public services in languages people best understand[1]. Not the local language or the default language of country of origin. Someone from South America might not actually speak Spanish. They might speak Quechua for instance.

Brings up the point, 'internationalization' as a concept is utterly broken. People speak languages not countries. Also very likely that the non default language users may also be most in need.

Probably should be a list of falsehoods programmers think about language.

[1] Local school district I think at one time listed 43 languages.

I use a VPN for all traffic, using an exit server in Finland. I don't speak a word of finnish.

I have chatted with people who work on internationalization about IP based detection. The explanation I've always heard is that `Accept-Language` is all but useless. For a significant number of users the value is, incorrectly, `en-US`. Because of how unreliable `Accept-Language` is, IP based detection is unfortuantely has a much lower false positive rate.

In general, I expect a web site, if it's available in more than one language, to offer me a menu where I pick the language I want. It's easy to do and it works for everyone. Usually the languages will also be accompanied by a little flag for extra non-ambiguity.

See the "Change region" menu at the top of https://www.nintendo.com, or the obvious "English" selector amidst the gibberish at the top of https://www.ikea.com/jp/ja/ .

Given how easy and robust this is... what exactly is the problem with using Accept-Language as your first guess? Maybe if more sites honored Accept-Language, users would start setting it.

You still have to pick a default region/language unless you want to force the user to make a selection before they get anywhere which is a mess UX-wise

Ikea still uses such a landing page if you go directly to ikea.com .

If it's specifically that value that's a problem then I'd suggest the following heuristic: Use geolocation if `Accept-Language` is `en-US`, otherwise respect the header.

Technical users in Switzerland/Japan/wherever who want to see English pages can set their header to "en-GB" or "en-AU" to avoid getting caught in the heuristic.

Edit: I'm sure it's much more complicated than that, of course- l10n isn't my job and I don't want to be this guy https://xkcd.com/1831/

Incorrectly... or not? I have my Accept-Language set explicitly to en-US, on purpose, in the small hope I stop getting localized results. I don't care about Polish Internet, I want results from the whole Internet.

But how can the Accept-Language header be useless? I suppose almost nobody sets this header by hand and it is derived from whatever environment your browser is in.

I mean if my Windows or my phone/tablet is set to German, showing me a German UI, wouldn't then the browser use that setting for Accept-Language?

I can't understand how this would produce unreliable results unless there were some serious bugs in browser support (hopefully in the past).

> Please just use English

Excuse me, sir. I would rather prefer it to always be German. Or French. Or Spanish. Or Italian.

Not everybody speaks English, so it's perfectly fine to try and guess the person's language first - and allow to change it later.

Also, once the UK has left the EU, English is not even going to be an official language here anymore - so I can see no reason why English should be the default in Europe.

Ireland isn't leaving, so English will still be an official language.

Not true - each European country can only choose one language - and Ireland chose Gaelic, since Britain chose English ;)

Yes, as an English speaker living in Germany, this is also something that drives me up the wall, but has also gotten better recently. My personal favorites, the cream of the crop:

* Netflix: full site in English, subtitles (and audio tracks) for all shows available in German, English and several other languages. Only a few edge cases where the source (written) material appearing on-screen has already been hard-translated in.

* Amazon: will let you turn on full English experience, up to and including...English speaking support (both written and verbal). This is truly the cream of the crop in terms of quality of experience.

I really appreciate any German-oriented site that offers English as a possible language in spite of knowing my location. And if somewhere on your servers you already have English available, you would be silly to not make that available world-wide (where it makes sense).

Of course, these are two of the biggest tech companies, so perhaps that speaks to the cost of internationalization.

pff this is nothing, German and English are quite similar languages, try living as foreigner in China, very enjoyable experience online /s

I live in Switzerland too and especially Youtube never seems to know whether to serve me ads in German, French or Italian (even though Google should know I almost exclusively consume English content). So I can sometimes watch several videos in a row, where I get served the same ad in a different language.

But don't ads on YouTube follow the more classic advertisement 'demographics', where location is more relevant than language? And it may just be cheaper and easier to just target all of Switzerland?

This is where I feel a bit better about targeted advertising, if they're not able (as in at least not cheaply enough to be worth it) to serve me an ad in a language I understand, they still have a long way to go.

And that's in Switzerland where people have a lot of disposable income and thus must be more attractive to advertisers than most places, despite being relatively small.

My original compilers had error messages in English, German, French, and Japanese. It was a fair amount of work to make that work. I was pretty proud of it.

Turns out that I couldn't find anyone who ever used the non-English messages. Everyone preferred the English messages, because everything else with programming was in English.

So I abandoned worry about that, and haven't heard anything about it since.

(Though the compiler uses UTF-8 strings for all text.)

One problem with the translated GCC messages was that GCC didn't use error numbers at the time (still doesn't, I think, I didn't notice anyway), so it was basically impossible to reference the error except by full error message.

MSVC on the other hand always includes a warning/error number, so there less issues talking about a particular error independent of language.

Since I gave up on translated error messages, using search and the error message string works fine (much better than a catalog of error codes).

Presumably you'd be fine with Russian too, but by the sound of it you've given up on ever seeing THAT. ;-)

(personally English is also often first choice, especially for tech documentation. Translations often seem to be an afterthought and can be pretty random. You need to translate randomly picked words back to the original jargon as best you can and hope you didn't miss anything. )

Yeah tech documentation is a really interesting edge case for translation, as a good chunk of it will be new vocab and/or simply invented vocab for most native English speakers reading it anyway, so it's actually a pretty philosophical question about language itself if/how that stuff can be translated to other languages at all.

It's because a huge number of people just discovered the net and they don't surf as we are used to, like they don't search as we do etc. I don't want to sound elitist, many of them don't have the experience of it. They are going to repeat all behaviors and fall in all traps since we were new on the net. I see this in the way my kids use their phones.

You are an edge case. Go to the site in English, remove the root URL from your history. So it always suggests the English version ( URL) of the website.

Best practise of a website ( eg. Google) don't suggest redirection based on an IP, which means a correct URL or cookie would suffice in most cases

The worse offender here is nba.com, which Geo-redirects you to some local TV channel based on X-Forwarded-For header.

Sounds more like you'd prefer an agreed upon lingua franca.

Reply on news.ycombinator.com