The solution on github seems to be Javascript-reliant, which can run into script-blocking issues.
First, I want to clarify that the prototype was made just to communicate the idea in terms of interaction. The The implementation is just a quick hack to simulate this interaction.
For a production implementation I can image the whole list of languages to be sent to the client, and then, the list being shortened by Javascript. For those users without Javascript (from screen-readers, to Search engine crawlers) the same list of links they receive now will be available for them.
For a GeoIP solution, this relies on good information about what languages
are relevant to GeoIPs. Do we have such a good set of data?
When we tested language selection on translatewiki.net, which uses the combination of browser accept language + previous choices + geo-ip, it worked well. Our data may not be perfect but it can be improved over time as we notice that some suggestions are missing.
In my opinion the critical points to verify are that:
- *Our predictions work most of the time.* We try to include minority languages at their region. This is something that some speakers of major languages may complain abouthttp://www.mediawiki.org/wiki/Talk:Universal_Language_Selector#x.22Common_languages.22_19010, but I agree with you that we still need to make them easily findable. - *For the cases when predictions fail, they fail only once* (the new language selected is considered for the suggestions in the future). - *It is clear that the suggested languages are not the only ones.* So that the user does not get the wrong impression that his language is lacking when it is not. Verifying this point is essential to ensure that some languages not appearing initially are not affected negatively.
I think that even in the edge cases (first-time users, with an misconfigured browser, accessing from a region where a language is not commonly spoken), it is easier to find it using the proposed language selector with the help of search than locating it in a plain list with 200 items. In any case, we need to keep verifying with users that the above premises hold true.
On Sat, Apr 20, 2013 at 9:58 AM, Shimmin suburbanpanther@yahoo.co.ukwrote:
A couple of thoughts occur to me.
The solution on github seems to be Javascript-reliant, which can run into script-blocking issues. I don't know what proportion of visitors might be using computers with script-blocking, and what proportion of those would think/know how to/have permissions to overcome it. Or using computers/browsers where JS is inclined to break. It might be completely minimal, but I thought it was worth mentioning.
For a GeoIP solution, this relies on good information about what languages are relevant to GeoIPs. Do we have such a good set of data? I'm thinking particularly of language communities outside their traditional homelands, Cantonese in Liverpool for example. Also, language density is a complicating factor. If you use a list of 10 languages based on GeoIP, then in some areas it's more than enough, while in others it's a fraction of the local languages. I'm not sure what the best way to overcome that one is.
I'm also concerned that a measure like this will tend to reinforce the dominance of major languages on the net. People will not necessarily take that extra step to check the language list just in case their language is on it, especially for lesser-wikified languages; adding an extra step always pretty much makes things more unlikely. I wonder whether the huge list we see at present encourages people to search for their own language, while a small list that doesn't immediately show it is less encouraging. The many wiki-readers who don't edit will presumably not have any preferences saved, so would potentially have to set their language choices every visit - or might simply not bother if it's unlikely to offer many articles anyway. So they would simply read the English/French/Russian articles, and the minority wikis would be further neglected and the language further undermined. This is obviously all speculation; I'd be interested to see any hard information on this. It's a different set of problems from those of interlang editors but one worth considering, particularly as you're talking about making this the default. Minority languages have a hard enough time as it is...
In terms of link ordering, it would perhaps make sense for articles related to a particular language to emphasise those links (either in a "Relevant to this article" section, or by formatting of some kind)? So articles on French people, things and places might highlight French - although of course there's other French languages to consider so that could get complicated.
--Shimmin
On 19/04/2013 10:05, സുനിൽ (Sunil) wrote:
I suggest the list of languages should be displayed according to the size/quality also
On Fri, Apr 19, 2013 at 3:43 AM, Yuri Astrakhan yastrakhan@wikimedia.orgwrote:
There are a few things (IMO) that should be done to langlist ordering:
- Group by alphabet
People who understand latin alphabet should get a list of all latin-using languages listed/sorted together. Cyrillic is a separate group, and so are various asian and middle-eastern languages. I have seen other sites do this (e.g. Google, but I can't quickly locate an example right now). Having all languages bunched up together make going through them extremely painful - one has to skip all the scripts not understood.
- Each wiki site has different ordering requirements - like Hebrew and
Hungarian wikis want English as the first link, or 'nn' uses 'no','sv','da' before all others. See pywikihttp://svn.wikimedia.org/svnroot/pywikipedia/trunk/pywikipedia/families/wikipedia_family.py- interwiki_putfirst
- Lastly, but IMO - most importantly, we should honor user settings or
browser settings. If my browser sends *Accept Language: en-US,en;q=0.8,ru;q=0.6*, it would be good to show english & russian at the top, followed by others.
All this can (and should) be done in javascript, without affecting servers.
And for historical reasons: bug 2867https://bugzilla.wikimedia.org/show_bug.cgi?id=2867... i filed it in 2005, it has over 60 votes (highest count in bugzilla if i'm not mistaken)...
--Yuri
On Thu, Apr 18, 2013 at 9:57 PM, Brion Vibber brion@pobox.com wrote:
I was traditionally in favor of keeping the full language list visible, but.... it's just too damn big in many cases and is hard to search through on any device. On touch devices it's difficult to pick a correct item from the list as all the links are adjacent (though if you zoom it's ok).
Definitely we need something improved, and if we're going to improve it we need to do it for the default or we're failing to serve 99% of our readers...
I'm not sure about the current demo; one thing that bugs me is that there's a very small tap/click target for getting the full language list call-out. Clicking on "Language" just hides/shows the short list, it doesn't do anything. Clicking the "settings" gear icon next to "Languages" brings up a call-out with language-related settings.... none of which help you get to another language version of the wiki.
On the mobile site we've collapsed the whole thing to an "Other languages" section or button (depending on if you're in beta mode) at the bottom of the article, and this seems to have gotten good usability responses from mobile users.
On Thu, Apr 18, 2013 at 12:47 PM, David Gerard dgerard@gmail.com wrote:
On 18 April 2013 20:43, David Gerard dgerard@gmail.com wrote:
On 18 April 2013 17:50, Pau Giner pginer@wikimedia.org wrote:
Please let me know if you see any possible concern with this
approach.
My first thought is of how upset people were when the first version
of
Vector hid the language links by default. I would suggest being sure there will be little or no similar objection.
(hit send too soon, sorry)
A simple solution that would avoid a similar reaction is: do not do this by default - make it only for logged-in users who want it that way.
Possibly for default users, you could put the heuristically-calculated likely preferred languages at the top. But keeping the rest of the list below, right there on display, will (I predict) be favoured, as advertising the many languages of Wikipedia is a strongly-held value of many Wikimedians.
- d.
Wikitech-l mailing list Wikitech-l@lists.wikimedia.org https://lists.wikimedia.org/mailman/listinfo/wikitech-l
Wikitech-l mailing list Wikitech-l@lists.wikimedia.org https://lists.wikimedia.org/mailman/listinfo/wikitech-l
Mediawiki-i18n mailing list Mediawiki-i18n@lists.wikimedia.org https://lists.wikimedia.org/mailman/listinfo/mediawiki-i18n
Mediawiki-i18n mailing listMediawiki-i18n@lists.wikimedia.orghttps://lists.wikimedia.org/mailman/listinfo/mediawiki-i18n
Mediawiki-i18n mailing list Mediawiki-i18n@lists.wikimedia.org https://lists.wikimedia.org/mailman/listinfo/mediawiki-i18n