402821 - Use apache rules instead of symlinks for language fallback

Status: RESOLVED FIXED

(www.mozilla.org :: General, defect)

Description

[Wil Clouser [:clouserw]]

On mozilla.com, the following locales:

'en','es','no','pt','sv'

are using symlinks to fallback to other locales.  We should be using apache redirects so applications looking at the filesystem don't think those directories actually exist.  This should also fix problems like bug 399106 since we'll be redirecting instead of rewriting.

Comment 1

[Reed Loden [:reed]]

Sounds good to me.

Comment 2

[Wil Clouser [:clouserw]]

Attachment: Add logic for redirection of locales

I saw we were already doing a similar thing with ja-JP-mac -> ja, so instead of using apache I just tweaked the prefetch script.  The patch doesn't remove the symlinks, but after applying it you should be able to remove them with no problems.

Requesting review from anyone.

Also the stat-keepers will be happy with this since people will actually get redirected when going to the mapped languages instead of just being shown the content.

Comment 3

[Wil Clouser [:clouserw]]

Added some people who have asked me about en vs. en-US in the stats before to keep them in the loop.

Comment 4

[Alex Polvi]

It would be great to have this change in ... would make our stats more consistent. 

Comment 5

[Reed Loden [:reed]]

Comment on attachment 297254 [details] [diff] [review]
Add logic for redirection of locales

...
>+/**
>+ * If we support a locale but want to redirect it to another, add it to this array.
>+ * Format:
>+ *      lower(requested_language) => mapped_language
>+ */
...
>+// If we're remapping the language to another
>+if (array_key_exists(strtolower($lang), $lang_remap)) {
>+    $lang = $lang_remap[strtolower($lang)];
> }

So, the patch as a whole is fine and will work, but I just find it somewhat odd how you're dealing with case in both the $lang_remap and in the above code. Could you give me an explanation as to why you're doing it this way? Just something seems off about it in how you have some stuff as lowercase while other stuff as the correct case. If there's a real need for all this case swapping, then that's fine, but it just seems fragile from a general overlook perspective. r=me with your reasoning. :)

Comment 6

[Wil Clouser [:clouserw]]

Everything is being lowercased while we're trying to detect what language to use, since there's no guarantee a browser won't send "en-uS".  At the end of the detection we make the format xx-YY (with correct case) and check if it's valid, if it is, we use it from there on.  Does that make sense?

Comment 7

[Wil Clouser [:clouserw]]

This is live in r11627