Best practice on URL design of a multi-language site

I've wondered if there's a best practice on defining the URL
structure of a multi-language site. I've seen the following
pattern, requesting:

https://addons.mozilla.org/

redirects me to:

https://addons.mozilla.org/en-US/seamonkey/

I suppose the preferred language is initially determined from the
'Accept-Language' request header. Next, wherever I navigate in the
site all URLs start with:

https://addons.mozilla.org/en-US/

e.g.:

https://addons.mozilla.org/en-US/seamonkey/addon/433

When I pass that URL to a person which has different language
preferences he will still be served the English variant which I
rather don't want. Providing URLs serving specific language
variants (e.g. by supplying 'alternate' LINKs in the HEAD of HTML
document) is o.k. but using "default" URLs serving a language
variant depending on the current visitor preferences is more
practical in my opinion.

Given other large sites like http://msdn.microsoft.com/ employ the
same "fixed" language URLs - what do you think is better, having:

http://www.example.net/article

serving the language variant as the current user preference is, or
having it always redirect to:

http://www.example.net/en-US/article

(or the whatever the current user preference is) which serves only
the given language variant?

--
Stanimir

On Sun, 25 May 2008, Stanimir Stamenkov wrote:

I've wondered if there's a best practice on defining the URL structure of a
multi-language site.

That depends on:

- whether you translate *all* pages in *all* available languages so that
you have never the situation that you have to serve a page in a
language the user understands but does not prefer

- whether you want to rely on the user to configure his browser according
to his real preferences; many users don't

If the answer to the first question is "yes", I would separate the
languages at the root level ("http://yoursite.com/en/..."), otherwise at
the leaf level ("http://yoursite.com/info/products/some_product_en").

In any case, the links in each page should point to other pages in the
*same* language so that the user's last choice overrides his general
preference. A link to the same page in other languages is very useful so
that the user can switch languages without losing his position in the
navigation.

And (not your question) be sure only to serve languages the Web author is
familiar with. Babelfish output is ridiculous and conveys no information.
If you have to offer something related to Web design, marketing, or
translation and internationalisation, grammar errors show lack of
competence and will not be tolerated, otherwise *minor* errors will be
tolerated.

--
Helmut Richter

Sun, 25 May 2008 16:20:50 +0200, /Helmut Richter/:

On Sun, 25 May 2008, Stanimir Stamenkov wrote:

I've wondered if there's a best practice on defining the URL structure of a
multi-language site.

That depends on:

- whether you translate *all* pages in *all* available languages so that
you have never the situation that you have to serve a page in a
language the user understands but does not prefer
[...]

If the answer to the first question is "yes", I would separate the
languages at the root level ("http://yoursite.com/en/..."), otherwise at
the leaf level ("http://yoursite.com/info/products/some_product_en").

What would you comment on my point this approach doesn't provide an
uniform URL which could serve one or other language variant specific
to the user's preference? That preference could be set through
browser setting and indicated in the 'Accept-Language' request
header and additionally overridden by site set cookie. Of course if
the user doesn't bother to set any of these appropriately he will be
served what best matches the server algorithm of choosing the
language variant. The 'Accept-Language' browser setting is often
pre-configured depending on the OS regional setting so it is very
likely it will indicate an actual user preference.

--
Stanimir

Scripsit Stanimir Stamenkov:

I've wondered if there's a best practice on defining the URL
structure of a multi-language site.

There's no specification as far as I know, but as regards to URLs of
language-specific versions, the best approach seems to be to make
language code one segment (between "/" characters) in the path, as in
the example you describe:

https://addons.mozilla.org/

redirects me to:

https://addons.mozilla.org/en-US/seamonkey/

I suppose the preferred language is initially determined from the
'Accept-Language' request header.

That would be my guess too, confirmed by some testing. By the way, I get
the en-US version when my Accept-Language header says just "en" or just
"en-AU", which is OK (but naive implementations might miss this), and
also when there is no such header, which is somewhat questionable, and
even when it says "zu" (for Zulu), which is simply wrong by the
protocol, though maybe not quite so in practice.

Next, wherever I navigate in the
site all URLs start with:

https://addons.mozilla.org/en-US/

Yes, because you were redirected, and the new page uses such explicit
links.

https://addons.mozilla.org/en-US/seamonkey/addon/433

When I pass that URL to a person which has different language
preferences he will still be served the English variant which I
rather don't want.

Understandable, but you need to know that it's a language-specific
version. The generic ("not language-negotiated") version is
https://addons.mozilla.org/seamonkey/addon/433

The site could be more explicit about such issues and should perhaps not
use redirection but simply serve the resource under the specified URL.
But then users would have the problem of finding out the
language-specific URLs.

Providing URLs serving specific language
variants (e.g. by supplying 'alternate' LINKs in the HEAD of HTML
document) is o.k. but using "default" URLs serving a language
variant depending on the current visitor preferences is more
practical in my opinion.

I would say that ideally the URLs should be generic but the pages should
contain information about the language-specific URL, for users who might
wish to use it. That is, e.g. the URL
https://addons.mozilla.org/seamonkey/addon/433 would be shared by
different language versions but each of them would carry (probably at
the end) a note like
"The URL for this specifically US English version for this page is
https://addons.mozilla.org/en-US/seamonkey/addon/433"
or something like that, naturally in the language of the document.

Both URLs might be needed e.g. when linking. If I have a page in Finnish
and I wish to refer to that page, I probably want to refer to the
Finnish version specifically - partly because I know that many people
with Finnish as first language use browsers with a factory setting of
Accept-Language: en-US and they don't even know this. But if my page
were in Zulu, I would probably use the generic URL, since I would not
know which of the available languages is each reader's preference (and,
besides, some day there might be a version of the page in Zulu, and
language negotiation might yield it when the generic URL is used).

--
Jukka K. Korpela ("Yucca")
http://www.cs.tut.fi/~jkorpela/

Scripsit Stanimir Stamenkov:

The 'Accept-Language' browser setting is often
pre-configured depending on the OS regional setting so it is very
likely it will indicate an actual user preference.

No, it is very likely that it is bogus, such as "en-US" or "en"
independently of the user's actual first language. Regional settings
cannot be relied on, and browsers need not (and often do not) consult
them.

Moreover, even if the browser setting is taken from the OS regional
setting and this setting reflects the user's primary language, the
Accept-Language header will mostly be _wrong_ in the essential sense of
not reflecting the user's language preferences.

Most people on Earth actually understand more than one language, at
least to some extent. Things might look different from the US
perspective, but even the US, quite a many people understand two or more
languages, at least to the extent of being able to understand something
about a web page.

For example, if you get a new computer in Finland, you will probably get
it with browser(s) configured to send Accept-Language: en or
Accept-Language: fi, partly depending on whether they get a
Finnish-language or an English-language version of the OS. (There are
many reasons why it may actually be an English version.) Both are just
wrong for most Finns - a typical _correct_ header would be
Accept-Language: fi, en, sv possibly followed by other codes (de,
fr,...).

But most people don't know about this. So the browser might ask for
English versions only (so that the user might never get a
language-negotiated version in his native language) or it might ask for
Finnish versions only (so that theoretically correct language
negotiation will usually result in a dialogue that just specifies the
available versions).

--
Jukka K. Korpela ("Yucca")
http://www.cs.tut.fi/~jkorpela/

On Sun, 25 May 2008, Jukka K. Korpela wrote:

No, it is very likely that it is bogus, such as "en-US" or "en" independently
of the user's actual first language. Regional settings cannot be relied on,
and browsers need not (and often do not) consult them.

Does anyone know of a real-life Website that has more than one regional
version of the same language, such as en-UK/en-US/en-IN/en-AU/en-ZA or
de-DE/de-AT/de-CH? For the German sublanguages, the major differences are
in judicial jargon, and then the contents, and not only the language, is
also different. (Well, there is a special Austrian vocabulary for regional
cuisine but using Austrian terms for Austrian dishes is vital even for a
de-DE audience, no matter whether they understand it.)

--
Helmut Richter

On Mon, 26 May 2008, Helmut Richter wrote:

Does anyone know of a real-life Website that has more than one
regional version of the same language,

Define "pt" or "pt-PT" or "pt-BR" as your preferred language
and visit http://www.google.com/webhp .

Or go directly to
http://www.google.com/webhp?hl=pt
http://www.google.com/webhp?hl=pt-BR
http://www.google.com/webhp?hl=pt-PT

You can try the same with Chinese (zh-CN, zh-TW).

--
Bugs in Internet Explorer 7
http://www.unics.uni-hannover.de/nhtcapri/ie7-bugs

"fixed" language URLs - what do you think is better, having:

http://www.example.net/article

serving the language variant as the current user preference is, or
having it always redirect to:

http://www.example.net/en-US/article

I'd prefer http://www.example.net/article




--

ø¤º°`°º¤ø,¸¸,ø¤º°`°º¤ø,¸¸,ø¤º°`°º¤ø,¸¸,ø¤º°`°º¤ø,¸ ¸,ø¤º°`°º¤ø,¸
http://www.HyperPublish.com Catalogs, CD and sites with 1 tool
http://www.EasyWebEditor.com Create a nice Web site with ease
http://www.1site.info A professional Website quickly
http://www.EBooksWriter.com Discover the artist inside you!
http://www.PaperKiller.com Manuals, HTMLHelp, CHM quickly
http://www.CdFrontEnd.com Create autorun CD presentations

Visual Vision - http://visualvision.com http://visualvision.it
leader in hypertext authoring [ASP members, ESC members]