Remark: This is second of four related posts. Please read others (I, II, III, IV) too. And please forward all comments and discussions to dev-mdc list at lists.mozilla.org.

Statistics of localization communities

We did some statistical analysis for MDC with using the crawled data of MDC Japan Project. Based data is not the same as the database of MDC itself, but the statistical results might be thought to have accidental error no more than one tenth.

General statistics

First, we can compare the content volumes in a form of pages’ count. We have many namespaces in MDC to separate their contents, but we can focus only on some of them in this statistics. So, I excluded some namespaces, such as template, user_talk and so on.

In this graph, vertical scale is shown as the full scale of English site. Almost all ‘user’ pages are categorized in ‘English’ language, but this is from a spec bug of deki that the system in default creates its user’s page even for spam accounts. So mostly these are for spam accounts.

The second graph is scaled without English site but localization sites maximized. We can say that most of the contents are belongs to ‘main’ namespace, and other namespaces including ‘main_talk’ or ‘user’ are minor and might not be used heavily by localization communities. So they might use other resources like e-mail lists, chat or forums in other sites for discussions on translated pages.

And we can call the localization languages with top 6 page counts as the ‘Most Edited Localization Languages’ – zh-cn, fr, ja, pl, es and ko.

Among the ‘main’ namespace, we can separate the pages into two categories – redirected pages and non-redirected pages.

Some redirected pages are created in the early time of devmo as link pages for outer resources, but most of all redirected pages are created by the Deki converter during the deki switch in 2008. We might define the difference between redirected and non-redirected page counts to be new created pages from 2008 to the current. (I don’t know why for -pl, but the non-redirected page count is a bit smaller than redirected page count in -pl.)

Pages updated in time ranges

Here, we would compare counts of updated pages in each time domain like 3 monthes periods.

(click to enlarge)

Here, graphs for all languages are shown. Blue vertical line shows the deki switch (the same for all following graphs). Updated pages in English wiki slightly decrease but even now there are high rate for updates. This might be mostly because that developers are forced to update MDC when added new features or modified behaviors of products, and also MoCo has document writting team for English MDC.

(click to enlarge)

Here, graphs for non-english languages are shown. With this graph, it is CLEAR that counts of updates in localizaed wiki are really decreased even in a factor of 10 to 100. This is a serious situation for localization teams.

Users updated MDC in time ranges

(click to enlarge)

(click to enlarge)

Here, two panels show users who updated the wiki page at least once for all- and non-English wikis. Count of contributors to English pages does not changed largely. Contributors for each localization community is about one tenth of English, and seems to be changed a little.

These graphs include casual contributors who fixed a minor typo or miss in the wiki, so we might focus on the most valuable contributors who updated the wiki frequently. So, we might call users with edits more than 50 updates (revisions but not pages) per 3 months period as a power editor. 50 updates per three months are not so large, as we can say it as 1 update per 2 days.

(click to enlarge)

(click to enlarge)

Here, two panels show for every languages and ‘Most Edited Localization Languages’. In most localization language communities, these power users seems to be stopped their actives on MDC.

Raw data in table format

You can also get raw data in table format.

Current status of each localization community

French team

The French localization community, which counted as much as three to five very active members and many occasional contributors just before the switch was also heavily disrupted. Most of these contributors just walked away and don’t want to hear about MDC anymore. Other communities like the long-standing XULfr haven’t been able to provide an alternative. Nowadays, localizing MDC is mostly seen as a joke to scare away willful new contributors, and few people are taking seriously the idea to start working at it again. We’re beginning to receive complaints from developers because most documentation pages are so out of date that they’d rather prefer to get the english content. Worse, the obsolete pages are not marked as such and developers have probably been misled about standards and what Gecko can do.

Japanese team

Japanese localization community for MDC had marked current MDC system using the DekiWiki as DEAD because of the hardness to translate or update the wiki contents. And also they announced to do some of their localization work not on MDC but their local community site — modest, such as localizing contents of the mozilla hacks site as well as some weekly or by-weekly newsletters like about:mozilla. Due to the problems of the current MDC including the serious performance problems, they are no longer able to force remaining contributors or the newbies to use or translate in MDC, and are hesitated to announce widely to earn new contributors. Of course, as well known in global MDC community, they did such translation projects first on MDC, but with a such serious decision, they switched all of them, which are not the contents of English MDC, to the other site.