Firefox 3.5 Language Coverage

Update: fixed a bug where I had confused Maltese and Malayalam. Firefox’s numbers improve further :-)

The languages to be included in Firefox 3.5 are now confirmed, and Microsoft has just released the last 5 additional languages for IE 8 on XP they said were pending. That’s three months after the original release. I guess they still need to learn from our team on how to simultaneously ship them all at once ;-P (And it’s worth noting quite a few of their languages are still Vista and/or 32-bit only. Should we count a language if it’s not available on their most common OS?)

Anyway, it’s time for an update on our plans for world domination. A reminder for those joining the party: I have a spreadsheet which combines the Internet population of a country with its first-language-spoken percentages to try and get an accurate idea of what languages the world Internet population speaks.

The headline is that of the latest releases of all the browser vendors, Firefox now has the highest percentage of world Internet population first-language coverage – 95.7%:

Browser Count Percentage
Firefox 3.0 59 92.4%
Firefox 3.5 68 95.7%
Opera 9.6 35 84.8%
IE 7 77 98.3%
IE 8 57 94.6%
Chrome 39 91.6%

The key change from the last release of the data is that I’ve got much more up-to-date figures for Internet population from InternetWorldStats (thanks very much to them). There are 28.5% more people on the net than in the last release of the statistics – an extra 300 million people. Almost every country’s net population has gone up, but they’ve gone up by different amounts, and so that has had a knock-on effect on the relative importance of different languages, and of the percentages of the world covered. So, for example, Firefox 3.0 “dropped” more than one percentage point from 93.5% to 92.4%, despite still covering the same languages it always did! As more of the developing world comes online, the goalposts are moving :-)

The stats have their own page, where you can find the spreadsheet, FAQs and a more detailed explanation. A few noteworthy points:

  • If you want a current browser in Esperanto, Welsh, Persian, Irish, Galician, Icelandic, Kurdish, Occitan, Romansh or Sinhala, Firefox is the only game in town.
  • We are releasing 3.5 in Turkish despite the Turkish localizer breaking his leg. That’s dedication.
  • Axel says we are looking at taking a Malay localization from Ubuntu. That would be another 0.72%. Seth says Azerbaijani is in progress, and it could be that some friends of build engineer Ben Hearsum are looking at Tagalog. Exciting times :-)

10 thoughts on “Firefox 3.5 Language Coverage

  1. Gerv, this sounds really great – kudos to all the localizers as well as L10n-drives and even developers who makes that possible!

    That said, once the FF 3.5 hype dies off a little bit, I might come back to you for any analysis of what coverage we are able to provide with SeaMonkey 1.1.x, the en-US-only alphas (just as a baseline) and the first in-sync-localized SeaMonkey release ever, the upcoming Beta 1. Even though the numbers will surely not look nearly as good as any of the coverages in your table, it would be nice if we can come up with some numbers that visualize where we are there.

  2. Kairo: If you send me a list of the languages you’ve got, I can get you stats in about 5 minutes :-)


  3. Good show!

    I do have a couple of minor quibbles…

    First, I’m not sure I’d really put Esperanto on that “only game in town” list. Pretty much everyone who speaks it is an enthusiast, kind of like the people who speak Klingon (albeit with different motivations). They’re all fluent in other languages, by necessity. (Notwithstanding the widely-repeated claim that there are “a thousand” native speakers, a claim that Lindstedt appears to have pulled out of his backside and for which he provides no citation, I’ve never seen ANYONE claim that there are real-world towns and villages of native speakers of Esperanto.) Providing an Esperanto localization for them to play with is cool and everything, but it doesn’t make the browser “the only game in town” for them in at all the same way that a Persian/Farsi localization would be for many Iranians (most of whom don’t have regular internet access so far, but, as you point out, that may change over time).

    Also, I strongly suspect that the numbers display more precision than they really have. 95.7% makes it sound like you’ve actually got a reasonably accurate count of how many internet users there are in, say, India. Since so many users share computers, I very much doubt a count with three significant digits is really possible. On top of that, multiplying the percentage of first-language speakers by the number of internet users implicitely assumes there’s no correlation between first language and internet use, which is likely invalid in some cases. (I’d definitely put money on there being a correlation between first language and internet use in Iran, for instance.)

    Still, attempting to track this is great, and I’m sure the effort you’ve put into it is very much appreciated. I sincerely doubt anyone has a way to calculate more reliable numbers than yours, and it’s important for the translation community to have, as you put it, goalposts. So despite the inherent difficulties, it’s a good work you’re doing. Your numbers are probably a pretty rough estimate, but it’s the best thing available and a good deal better than not *having* an estimate.

    Purely out of curiousity, what languages does IE have that Firefox lacks?

  4. Jonadab: what I said was “If you want a current browser in Esperanto, Firefox is the only game in town.” I agree that there would be very few people who need a current browser in Esperanto, but that’s not the same thing.

    As for precision, I already knocked a digit off my spreadsheet’s number. 2 digits of precision would be, I think, a bit thin, three digits is arguably too precise. Do we really serve more first-language speakers than IE? A statistician would plug in all the data and its probabilities, standard deviations and so on and tell you that there’s a 78.3% chance that we do, but I’m happy just to use 3 digit precision. After all, if they beat us on this measure (which they could do by e.g. shipping a Persian localization) then I’d take the medicine like a man.

    You are right about assuming there’s no correlation between language and Internet use, and that being a flaw. But, as you say, there are no better stats until countries start adding “do you have access to the Internet” to the census, and releasing breakdowns by language. Given that many countries only do a census every 10 years, it’ll be a little while…

    For languages IE 8 has that Firefox lacks, see the spreadsheet :-) Bosnian, Konkani, Kyrgyz, Malay, and Uzbek.

  5. Bzzzt!

    Feel the Shiretoko Shock! So what am I personally most excited about in the new Firefox? Well, I have no uses for Private Browsing Mode – I think porn is a terrible corruption of God’s design for sexual relationships, my personal medical condition that…

  6. Gerv Status 2009-07-03

    This Week Governance Careful comments on fligtar’s proposal for changing the AMO sandbox model Started looking at the issue of whether we have private mailing lists which shouldn’t be private Worked with harvey on streamlining procedure for e-signing C…

  7. What about Latin? There are many people studying classic languages. Just for fun. And it would attract some media attention.