Language Analysis for FF3
Gervase Markham
http://weblogs.mozillazine.org/gerv/archives/2008/06/firefox_3_language_coverage.html
Gerv
pascal
> The results of my language analysis for Firefox 3 are available:
> http://weblogs.mozillazine.org/gerv/archives/2008/06/firefox_3_language_coverage.html
>
> Gerv
Bogus.
"They have at least 77 packs, we only do 48,"
No, we offer 48 fully localized builds of Firefox 3 and they offer 24
fully localized builds of IE7.
A language pack is not the same as a fully localized product.
If you want to count language packs in addition to fully localized
builds, then you should add the 28 langpacks we currently offer on AMO
and we get to a total of 76 locales supported versus 77 for IE.
Furthermore, microsoft and mozilla have a greatly diverging definition
of what a langpack is :
"The Internet Explorer 7 LIP is based on MUI technology. The Internet
Explorer 7 LIP translates a reduced set of UI elements and provides
approximately an 80 percent native language user experience"
80% of the UI done would not be accepted as a final firefox version
promoted on mozilla.com, but would be accepted on AMO where we have
several of the locales you said we don't have like Hindi:
https://addons.mozilla.org/fr/firefox/browse/type:3
Pascal
Seth Bindernagel
----- "pascal" <pascal....@free.fr> wrote:
These are very valid points that Pascal argues. Gerv, can you respond?
Additionally (or perhaps it's more to what Pascal is saying), I'd like to read how you define the term "Language Pack" in your post. We ship locales that are deeply integrated in our build and release system. As Pascal mentions, we also host language packs on AMO. To me, these are two distinctly different offerings that together provide many language offerings for our users. I think your post is a bit misleading...that or I am confused.
Michael Wolf
> If you want to count language packs in addition to fully localized
> builds, then you should add the 28 langpacks we currently offer on AMO
> and we get to a total of 76 locales supported versus 77 for IE.
Do you include langpacks for FF 2 here? If yes, you can add my private
ones for Upper Sorbian (hsb) and Lower Sorbian (dsb). For both are
packed builds (zip, tar.bz2), too. So you will have 78 locales, 1 more
than IE :-)
Gervase Markham
> Do you include langpacks for FF 2 here? If yes, you can add my private
> ones for Upper Sorbian (hsb) and Lower Sorbian (dsb). For both are
> packed builds (zip, tar.bz2), too. So you will have 78 locales, 1 more
> than IE :-)
Yes, in the "All localizations found" statistic.
Are you able to tell me in which country these two languages are spoken,
and what percentage of the population speaks them as a first language?
Gerv
Gervase Markham
> Bogus.
Whoa, easy, dude. It's just loose language, not some attempt to be rude
to people.
> A language pack is not the same as a fully localized product.
OK, fair enough. I was attempting to simplify and ask the question "is
it possible for this person to have a version of this product with the
UI in their language?". So I lumped together addons, packs, full builds,
MUI packs and all other sorts of things into one. Yes, they offer
somewhat different experiences, but any of them is far, far better than
having the UI in another language.
> If you want to count language packs in addition to fully localized
> builds, then you should add the 28 langpacks we currently offer on AMO
> and we get to a total of 76 locales supported versus 77 for IE.
This is what the "total localizations found" line is. So why is that
number not 28 greater than the FF3 number? A few reasons:
Firstly, I only count one version of English - so for example, English
(South African) isn't counted as separate. Again, this process involves
some simplification. I'm looking at percentage coverage, not absolute
numbers of languages supported, and a South African English person can
get by pretty well with American English (just as I, a British English
person, often get by with American English products). I've also
collapsed the two versions of Tamil although, knowing less about that
language, I'm less sure how reasonable that is.
Secondly, that page offers some packs which are also available as full
builds, such as Brazilian Portuguese. So that's not additional either.
Esperanto is a difficult case, because it is not the language of a
particular country or region, and so doesn't fit well with the
country-based method I'm using. Every method has flaws. There's an
apology to the Esperantists on the first page of the spreadsheet,
although I didn't put one in the blogpost.
Lastly, there are several African languages which we have packs for that
I now realise I haven't yet managed to integrate into the spreadsheet. I
apologise, and I agree that this needs doing. They are: Malagasy,
Ndebele (South), Northern Sotho, Siswati, Southern Sotho, Venda and
Wolof. These should also be added to the list of languages which only we
support. But they will make very little difference to the percentage
figures, because the internet populations of the countries in which they
are spoken is so small.
It's all a work in progress.
> Furthermore, microsoft and mozilla have a greatly diverging definition
> of what a langpack is :
>
> "The Internet Explorer 7 LIP is based on MUI technology. The Internet
> Explorer 7 LIP translates a reduced set of UI elements and provides
> approximately an 80 percent native language user experience"
That quote is from here:
http://www.microsoft.com/globaldev/DrIntl/faqs/lipfaq.mspx
and talks about Windows XP Professional Language Interface Packs (LIP),
not specifically IE. So you aren't comparing like with like. That's 80%
of the strings across the whole OS, noy 80% of IE.
As you will know, a large proportion of translation work for software
relates to error messages and other strings which are seen very rarely.
So (my conjecture is) a LIP doesn't mean that 4 out of 5 IE menu items
will be translated, it means that a whole bunch of obscure OS errors
will come up in e.g. Spanish rather than Catalan.
Quoting further from that page:
"Components not localized include Help and Support Center,
administrative tools, system error messages, event logs, Messenger, and
others. User Assistance (Help) files are not localized, and most
packages add one special localized topic to Help and Support Center."
So I agree that there is the question of what happens with Help. What is
the current position on translating Help? Is it required for a full
build? For an official pack (bottom of the download page)? For
addons.mozilla.org?
If someone has installed such a pack and finds that, in fact, core UI
items in IE are not translated, then I'll reconsider. But based on
available evidence, I think my logic is good here, and that my
simplification is not unreasonable.
Gerv
pascal
>
> If someone has installed such a pack and finds that, in fact, core UI
> items in IE are not translated, then I'll reconsider. But based on
> available evidence, I think my logic is good here, and that my
> simplification is not unreasonable.
>
> Gerv
I think your oversimplification is totally unreasonable and shows a lack
of understanding of our own localization process.
Pascal
Gervase Markham
> Lastly, there are several African languages which we have packs for that
> I now realise I haven't yet managed to integrate into the spreadsheet. I
> apologise, and I agree that this needs doing. They are: Malagasy,
> Ndebele (South), Northern Sotho, Siswati, Southern Sotho, Venda and
> Wolof. These should also be added to the list of languages which only we
> support. But they will make very little difference to the percentage
> figures, because the internet populations of the countries in which they
> are spoken is so small.
OK, so there's two things wrong with that paragraph.
Firstly, I had included Northern Sotho and Southern Sotho, just under
different names. So that's OK. Secondly, it turns out it does make some
difference. Once I've got any other feedback that there is, I'll publish
updated figures.
As I said, a work in progress :-)
Gerv
Gervase Markham
> I think your oversimplification is totally unreasonable and shows a lack
> of understanding of our own localization process.
Pascal, with respect, I really don't think that's true. For a start, the
question is about whether Microsoft's packs are as useful to an end-user
as ours are. So it's as much about their processes as ours.
Now let's say that you are right, and that their stuff is not as good as
ours for those LIP packs. Even if that's true, it's way, way better than
having 100% of the UI in your second or third language. What I'm doing
is effectively taking a 90% solution and a 100% solution, putting them
in the same basket, and contrasting them with the 0% solution (no
localization at all). I really don't think that's unreasonable.
What would you prefer? Do you think my spreadsheet should say that
Microsoft has no support at all for Afrikaans, Albanian, Armenian, ...,
Welsh and Zulu (all the LIP packs)? I personally don't think that would
be a fair reflection of the truth. But I'll add a line in for that if
you like.
<edits spreadsheet>
OK. IE with no LIP is at 85.7%, with 32 localizations. (But note that
some of the other figures have moved a bit, so wait till I publish the
sheet again for a comparison.)
Gerv
Gen Kanai
> pascal wrote:
>> Bogus.
>
> Whoa, easy, dude. It's just loose language, not some attempt to be
> rude
> to people.
Gerv,
With all due respect, I hope you can see that "loose language" to you
would be rude to the people who are tirelessly working to support our
l10n communities.
>> If you want to count language packs in addition to fully localized
>> builds, then you should add the 28 langpacks we currently offer on
>> AMO
>> and we get to a total of 76 locales supported versus 77 for IE.
>
> This is what the "total localizations found" line is. So why is that
> number not 28 greater than the FF3 number? A few reasons:
>
> Firstly, I only count one version of English - so for example, English
> (South African) isn't counted as separate. Again, this process
> involves
> some simplification. I'm looking at percentage coverage, not absolute
> numbers of languages supported, and a South African English person can
> get by pretty well with American English (just as I, a British English
> person, often get by with American English products). I've also
> collapsed the two versions of Tamil although, knowing less about that
> language, I'm less sure how reasonable that is.
The percentage coverage number is one metric that is useful but it
should not be the main metric that we judge our l10n efforts on.
> Secondly, that page offers some packs which are also available as full
> builds, such as Brazilian Portuguese. So that's not additional either.
Completely disregarding the efforts of the en-GB, en-SA, pt-BR, etc.
teams is not a great way engender support for your effort here.
> Lastly, there are several African languages which we have packs for
> that
> I now realise I haven't yet managed to integrate into the
> spreadsheet. I
> apologise, and I agree that this needs doing. They are: Malagasy,
> Ndebele (South), Northern Sotho, Siswati, Southern Sotho, Venda and
> Wolof. These should also be added to the list of languages which
> only we
> support. But they will make very little difference to the percentage
> figures, because the internet populations of the countries in which
> they
> are spoken is so small.
Again, your focus on percentage of usage is, in my opinion, a poor
metric for a number of reasons. It's more important that people can
get Firefox in languages that Microsoft will never support like the
11 languages of South Africa that Dwayne and the Translate.org.za
folks have helped make a reality. In many ways it is the smallest
communities which are the most important as they are the most
marginalized- by the percentage metric as well.
Microsoft will never release a support for Dzongkha due to political
pressures. Therefore Bhutan's Department of Information and
Technology developed their own version of Linux (with Firefox) in
their own language. Please add Dzongkha to your list if you do not
have it on there already.
http://dzongkha.sourceforge.net/
> It's all a work in progress.
Fair enough. That could have been said up front more clearly.
> So I agree that there is the question of what happens with Help.
> What is
> the current position on translating Help? Is it required for a full
> build? For an official pack (bottom of the download page)? For
> addons.mozilla.org?
Help is now online with SUMO, so it is a separate effort.
> If someone has installed such a pack and finds that, in fact, core UI
> items in IE are not translated, then I'll reconsider. But based on
> available evidence, I think my logic is good here, and that my
> simplification is not unreasonable.
If you are asking for feedback, and get feedback from the person who
manages our global l10n effort, and then disregard that information,
I hope you can see why there would be consternation. If you are
looking for support for your efforts, especially from Mozilla's own
l10n community, I humbly suggest that you reconsider your
simplifications.
Gen Kanai
Peter Weilbacher
If I may chime in, they are both languages spoken by a small population
in Germany. Wikipedia says it's 55000 and 14000 people for Upper and
Lower Sorbian, respectively, so less than 0.1% of German inhabitants.
Peter.
Michael Wolf
> Michael Wolf wrote:
> Yes, in the "All localizations found" statistic.
>
> Are you able to tell me in which country these two languages are spoken,
> and what percentage of the population speaks them as a first language?
AS Peter wrote in Germany, precisely in Eastern Germany in the German
federal countries Saxony and Brandenburg, see:
http://en.wikipedia.org/wiki/Sorbian_languages
http://en.wikipedia.org/wiki/Upper Sorbian
http://en.wikipedia.org/wiki/Lower Sorbian language
The number of speakers Peter mentioned could be right. I don't know the
exact numbers, I heard about 40,000 (Upper):20,000 (Lower).
Both are autochtoneous West Slavic languages, that means that Upper and
Lower Sorbs aren't ethnic minorities of other nationalities. Danes are
e.g. a non-autochtoneous minority in Germany. Another autochtoneous
languages are Northern Frisian and Eastern (Saterlandic) Frisian which
are spoken in Germany only.
Upper Sorbian is spoken in the Saxon region Upper Lusatia, in a triangle
of the towns Kamenz, Hoyerswerda and Bautzen (to teh north east resp.
east of Dresden)
Lower Sorbian is spoken in the Brandenburg region Lower Lusatia around
the city Cottbus.
Michael
Robert Kaiser
> Again, your focus on percentage of usage is, in my opinion, a poor
> metric for a number of reasons. It's more important that people can get
> Firefox in languages that Microsoft will never support like the 11
> languages of South Africa that Dwayne and the Translate.org.za folks
> have helped make a reality. In many ways it is the smallest communities
> which are the most important as they are the most marginalized- by the
> percentage metric as well.
This also applies to Michael's efforts on the Sorbian languages, or even
Esperanto, by the way. People native to languages like Sorbian usually
speak another language (like German in this case) fluently, and they are
small groups, so a for-profit endeavour like MS's will never support
them in the way our contributors can.
The question is if we statistics are about "how much of the population
can we reach?" or about "how many people do we support with software in
their own native language?"
The answers to those questions are a lot different.
Robert Kaiser
Gervase Markham
> The percentage coverage number is one metric that is useful but it
> should not be the main metric that we judge our l10n efforts on.
Sure. Who said it was? Not me.
>> Secondly, that page offers some packs which are also available as full
>> builds, such as Brazilian Portuguese. So that's not additional either.
>
> Completely disregarding the efforts of the en-GB, en-SA, pt-BR, etc.
> teams is not a great way engender support for your effort here.
I'm not "completely disregarding their efforts"! I'm a Brit - I use the
British English localization, for goodness sake! I'm just trying to take
a higher-level view.
Now if British English and South African English were mutually
unintelligible, then that would be a fair criticism. But they aren't.
Microsoft don't ship en-SA. Does that mean they have no support for
English speakers in South Africa? I don't think it's fair on them to
claim that.
>> Lastly, there are several African languages which we have packs for that
>> I now realise I haven't yet managed to integrate into the spreadsheet. I
>> apologise, and I agree that this needs doing. They are: Malagasy,
>> Ndebele (South), Northern Sotho, Siswati, Southern Sotho, Venda and
>> Wolof. These should also be added to the list of languages which only we
>> support. But they will make very little difference to the percentage
>> figures, because the internet populations of the countries in which they
>> are spoken is so small.
>
> Again, your focus on percentage of usage is, in my opinion, a poor
> metric for a number of reasons.
I think it's the best we are going to get of a measure of the absolute
number of people which we serve in their native language. I agree it's
not the only metric, but it's a useful one. It also gives us some guide
as to where we might focus our efforts next, to help the maximum number
of people.
> It's more important that people can get
> Firefox in languages that Microsoft will never support like the 11
> languages of South Africa that Dwayne and the Translate.org.za folks
> have helped make a reality.
I don't think it's easy to make comparisons like that; how do you say
whether it's more important to give the power of Firefox to 5 million
Bengali speakers, or to give a native-language browser to 89,000 Venda
speakers, many of whom probably speak another language as well? I don't
think the answer to that question is obvious - or useful, for that matter.
> Microsoft will never release a support for Dzongkha due to political
> pressures. Therefore Bhutan's Department of Information and Technology
> developed their own version of Linux (with Firefox) in their own
> language. Please add Dzongkha to your list if you do not have it on
> there already.
>
> http://dzongkha.sourceforge.net/
OK. To add it, I need to know a few things.
Is a Firefox localization available on its own? Or is Firefox in
Dzongkha only available to Linux users who install that distribution? I
don't think it's fair to mark it as supported if you need to install an
entire OS that you weren't using before.
What percentage of the people of Bhutan speak Dzongkha as their first
language? (I can try and look this up, but if you have good information,
let me know.)
>> It's all a work in progress.
>
> Fair enough. That could have been said up front more clearly.
Well, I did post in this group two weeks ago asking for help. What else
would you have liked me to have done? Now I've posted some results, lots
of people are providing more information.
>> So I agree that there is the question of what happens with Help. What is
>> the current position on translating Help? Is it required for a full
>> build? For an official pack (bottom of the download page)? For
>> addons.mozilla.org?
>
> Help is now online with SUMO, so it is a separate effort.
OK. In that case, I think it's fine, in my approximation, to mark MUI
builds down as that language being "supported" by IE. Because they don't
have help translated, but we don't require it either.
> If you are asking for feedback, and get feedback from the person who
> manages our global l10n effort, and then disregard that information, I
> hope you can see why there would be consternation.
I'm happy to make a distinction between IE's "full" localizations and
the MUI ones, and the current spreadsheet has entries for both. But
Pascal has not yet given any concrete reason why MUI builds are not as good.
Gerv
chris hofmann
The later question is a lot more interesting to me, and beyond that the
question "how many more *could* we reach by the addition of each new
incremental locale addition?" should get more attention.
That would help to set up a priority list and discussion of what new
locales would be most valuable to add.
Hindi, es-mx, and others are great places where I think we could fill
gaps to improve the experience for users with new locales or "tuned"
versions that better fit the needs of users globally, and in specific
areas, but we could use more analysis to back up those ideas.
From that we could recruit and help organize those locales to get off
the ground and shipping for upcoming releases.
Gerv, could your research be restructured to help answer those questions.
-chofmann
> The answers to those questions are a lot different.
>
> Robert Kaiser
> _______________________________________________
> dev-l10n mailing list
> dev-...@lists.mozilla.org
> https://lists.mozilla.org/listinfo/dev-l10n
>
Gervase Markham
> If I may chime in, they are both languages spoken by a small population
> in Germany. Wikipedia says it's 55000 and 14000 people for Upper and
> Lower Sorbian, respectively, so less than 0.1% of German inhabitants.
I've added both. Well done to Michael for his hard work :-)
However, and I hope I can say this without offending either of Peter and
Michael, if we now celebrate because we've got more packs than IE, we're
missing part of the picture.
When those numbers are scaled down by the German internet population
percentage, these two packs serve around 34,000 people, I suspect many
of whom speak German. Some of the packs we are missing, e.g. Bengali,
serve up to 5 million internet citizens, many of whom speak only that
language.
I agree that my spreadsheet is not the only metric. But number of packs
is not the only metric either. (Can we please discuss what good metrics
are calmly, without anyone accusing anyone else of denigrating someone's
work?)
Gerv
Seth Bindernagel
----- "Gervase Markham" <ge...@mozilla.org> wrote:
| Gen Kanai wrote:
| > The percentage coverage number is one metric that is useful but it
| > should not be the main metric that we judge our l10n efforts on.
|
| Sure. Who said it was? Not me.
|
| >> Secondly, that page offers some packs which are also available as
| full
| >> builds, such as Brazilian Portuguese. So that's not additional
| either.
| >
| > Completely disregarding the efforts of the en-GB, en-SA, pt-BR,
| etc.
| > teams is not a great way engender support for your effort here.
|
| I'm not "completely disregarding their efforts"! I'm a Brit - I use
| the
| British English localization, for goodness sake! I'm just trying to
| take
| a higher-level view.
|
| Now if British English and South African English were mutually
| unintelligible, then that would be a fair criticism. But they aren't.
|
| Microsoft don't ship en-SA. Does that mean they have no support for
| English speakers in South Africa? I don't think it's fair on them to
| claim that.
One interesting thing to consider is the other elements that go into making a localizations...
For instance, all of our localizers work with Mozilla to find the best localized search partners, newsfeeds, and RSS handlers. Imagine if you were using an en-US build in India, but kept receiving U.S.-centric default options. As a user, you might love to have a localized product that actually aligned with the search options, news feeds, and RSS handlers that were local to your community. See what I am getting at?
It doesn't have to be repeated to this crowd, but localization can often be about more than just translation. And, one could make a strong argument that en-GB, en-US, en-CA, etc. do provide a different experience for users. In addition to all that, there are colloquialisms, grammar, and spelling differences within those languages that localizers are working on. Those efforts should either be acknowledged in a different study or included in this one and described why that hard work by those localizers is worthy of inclusion.
(Now, some will probably have good comments to add about why creating localizations for every community might have its challenges...probably another very legitimate conversation to have one day.)
Overall, Gerv, I think you're on to something with this study. It's very cool to see how much coverage in the world's speaking population Mozilla covers.
But, I really think it's important to separate the work of every localization team and count it as a localization. As has been mentioned by Gen and Pascal, our localizer community works so tirelessly to do what they do. I am happy to discuss and collaborate with you on your study methodology...might be very interesting to refine how we are looking at it and then repost something.
-seth
irc nick: sethb (irc.mozilla.org)
Gervase Markham
> One interesting thing to consider is the other elements that go into
> making a localizations...
Agree on all points.
However, I suggest that all these things are secondary to having the
browser UI in your language at all. As I said to Pascal, what I'm doing
is taking 80%, 90% and 100% solutions and putting them together and
contrasting them with 0% solutions. I don't think that's absolutely
unreasonable.
It's not saying that the 80% solution and the 100% solution are
equivalent in all ways, only that both are way better than a 0% solution.
> Those efforts should
> either be acknowledged in a different study or included in this one
> and described why that hard work by those localizers is worthy of
> inclusion.
I'm not sure how this particular study could include them, unless we
adopted some way of rating a pack. If you do a full en-GB translation
but have en-US search engines, does that count as 90% serving the UK
public? And is that the same as serving 90% of the public 100% well? I
don't think one can really do that sort of metric meaningfully.
> But, I really think it's important to separate the work of every
> localization team and count it as a localization. As has been
> mentioned by Gen and Pascal, our localizer community works so
> tirelessly to do what they do. I am happy to discuss and collaborate
> with you on your study methodology...might be very interesting to
> refine how we are looking at it and then repost something.
Well, I could separate en-US, en-GB and en-SA, but I think it would only
be fair to say that Microsoft's English pack supported them all. And if
I did, it would have the following two effects:
- The percentage figures would stay _exactly_ the same
- Microsoft would appear to have created more language packs than it has.
So I'm not sure modifying the spreadsheet in that way would improve its
clarity.
Gerv
Gen Kanai
On Jun 14, 2008, at 1:22 AM, Gervase Markham wrote:
>> Microsoft will never release a support for Dzongkha due to political
>> pressures. Therefore Bhutan's Department of Information and
>> Technology
>> developed their own version of Linux (with Firefox) in their own
>> language. Please add Dzongkha to your list if you do not have it on
>> there already.
>>
>> http://dzongkha.sourceforge.net/
>
> OK. To add it, I need to know a few things.
>
> Is a Firefox localization available on its own? Or is Firefox in
> Dzongkha only available to Linux users who install that
> distribution? I
I have not seen it outside of the Dzongkha Linux distribution.
> don't think it's fair to mark it as supported if you need to
> install an
> entire OS that you weren't using before.
I don't follow your logic. There is no other OS which supports
Dzongkha, so to use a Dzhongkha Mozilla browser, you have to use that
Linux distribution. Whether it is available separately or not is not
relevant.
> What percentage of the people of Bhutan speak Dzongkha as their first
> language? (I can try and look this up, but if you have good
> information,
> let me know.)
Quickly, Wikipedia says first language of 130,000 and second language
of ~470,000.
http://en.wikipedia.org/wiki/Dzongkha_language
How accurate that is, your guess is as good as mine.
chris hofmann
> The later question is a lot more interesting to me, and beyond that
> the question "how many more *could* we reach by the addition of each
> new incremental locale addition?" should get more attention.
>
> That would help to set up a priority list and discussion of what new
> locales would be most valuable to add. Hindi, es-mx, and others are
> great places where I think we could fill gaps to improve the
> experience for users with new locales or "tuned" versions that better
> fit the needs of users globally, and in specific areas, but we could
> use more analysis to back up those ideas.
>
> From that we could recruit and help organize those locales to get off
> the ground and shipping for upcoming releases.
>
> Gerv, could your research be restructured to help answer those questions.
>
> -chofmann
>> The answers to those questions are a lot different.
>>
>> Robert Kaiser
Ricardo Palomares Martínez
> The results of my language analysis for Firefox 3 are available:
> http://weblogs.mozillazine.org/gerv/archives/2008/06/firefox_3_language_coverage.html
Huh, am I the only one _not_ seeing Gerv's post as a direct attack to
Mozilla's L10n community (even though I'm an es-ES localizer)? :-) To
me, what he is trying to find out is how many Internet population can
use Firefox vs. IE feeling that they are using the product in their
language.
Of course that IE "langpacks" will likely be of inferior quality or
less complete than Firefox ones. That will be true even for full
supported languages; after some time being involved in Mozilla L10n, I
found out some wrong translations to Spanish in Microsoft OSes. But,
after all, this is how Microsoft has achieved his position: they don't
try to do the best ever possible product, but the one "good enough"
for most people. And, let's face it, for a bengali person it will be
better to have a "not perfectly translated to bengali" product than an
en-US one. At least, this how I read Gerv's post.
Of course, another completely different question is: does this mean
Mozilla should pay bengali translators just because nobody is
interested to volunteer to do it? If so, how will all the not paid,
volunteer translators to the current XX (45, 48, 76, whatever)
languages feel about such decission?
Or maybe the real point is what can Mozilla do to attract people from
those countries without Mozilla localizations to volunteer?
JM2C
Michael Wolf
> I've added both. Well done to Michael for his hard work :-)
Thank you very much. :-)
>
> However, and I hope I can say this without offending either of Peter and
> Michael, if we now celebrate because we've got more packs than IE, we're
> missing part of the picture.
>
> When those numbers are scaled down by the German internet population
> percentage, these two packs serve around 34,000 people, I suspect many
> of whom speak German. Some of the packs we are missing, e.g. Bengali,
> serve up to 5 million internet citizens, many of whom speak only that
> language.
All Sorbs speak German. IMHO metric for localization statistics
shouldn't be based upon comparison with other software. In German we say
"der Vergleich hinkt", the comparison "goes lame", that means you
compare different things that are not really comparable.
The Sorbian languages are dying languages, they should be preserved. I
translated software into the Sorbian languages that they can meet the
modern requirements of computing. Otherwise these languages would
perhaps stagnate on a folkloristic level and sooner or later they would
be extinct. BTW, I'm not a Sorb, I'm an ethnic German.
The great advantage of Open Source, all Mozilla programs are based upon,
that they consider small languages because there is a community of
volunteers who do this work to realize certain aims and not to gain
money. They don't ask if it is worthwhile. They want to do it and
therefore they do it.
We must create locales both for small languages and for big "exotic"
languages. We should do one of this without leaving the other alone.
Long time ago (one or two years) Axel started a thread in this group
about the target of 100 locales for Fx. Unfortunately that thread ended,
if I remember well, in a discussion what localizations tools would be
better, simple editor, Mozilla Translator, translate toolkit etc. That
is secondary.
Michael
Simos Xenitellis
> Gervase Markham escribió:
>
>> The results of my language analysis for Firefox 3 are available:
>> http://weblogs.mozillazine.org/gerv/archives/2008/06/firefox_3_language_coverage.html
>>
>
>
> Huh, am I the only one _not_ seeing Gerv's post as a direct attack to
> Mozilla's L10n community (even though I'm an es-ES localizer)? :-) To
> me, what he is trying to find out is how many Internet population can
> use Firefox vs. IE feeling that they are using the product in their
> language.
>
What I like to see is constructive criticism.
I suppose it is the language barrier that causes troubles.
If ones reads your first two lines in the reply, they might think that
what you are trying to say is "Gerv's post is a direct attack to the
L10n community".
This is because some readers might be primed from the first replies that
Gerv received,
and think negatively. I misread it like that.
> Of course that IE "langpacks" will likely be of inferior quality or
> less complete than Firefox ones. That will be true even for full
> supported languages; after some time being involved in Mozilla L10n, I
> found out some wrong translations to Spanish in Microsoft OSes. But,
> after all, this is how Microsoft has achieved his position: they don't
> try to do the best ever possible product, but the one "good enough"
> for most people. And, let's face it, for a bengali person it will be
> better to have a "not perfectly translated to bengali" product than an
> en-US one. At least, this how I read Gerv's post.
>
> Of course, another completely different question is: does this mean
> Mozilla should pay bengali translators just because nobody is
> interested to volunteer to do it? If so, how will all the not paid,
> volunteer translators to the current XX (45, 48, 76, whatever)
> languages feel about such decission?
>
> Or maybe the real point is what can Mozilla do to attract people from
> those countries without Mozilla localizations to volunteer?
>
The importance of Gerv's post for me is that he put together information
that is lying around.
This information can then be disseminated, etc.
For languages that no translation project exists, there can be
a call for volunteers from Mozilla and local user groups to join up.
For languages that lang packs exist, but are not part of the official
Firefox,
it's an issue of getting in contact with the local teams and asking them
how to help
to get the work in Firefox.
Simos
Axel Hecht
While I agree that % of internet population covered is an interesting
metric, it doesn't give us any idea on where to focus our efforts.
Small chinese cities will outnumber most of the remaining languages in
terms of internet population, I guess. At least when you consider
"small" to be in chinese proportions.
I can imagine several more or less arbitrary metrics for where to focus,
all of which are some combination of population, internet population,
and the growth of the two.
I'd still add an estimate for the return of investment, which is the
thing we're struggling with the most.
I'll follow up on "recruiting" in a reply to Chris' post.
>> It's more important that people can get
>> Firefox in languages that Microsoft will never support like the 11
>> languages of South Africa that Dwayne and the Translate.org.za folks
>> have helped make a reality.
>
> I don't think it's easy to make comparisons like that; how do you say
> whether it's more important to give the power of Firefox to 5 million
> Bengali speakers, or to give a native-language browser to 89,000 Venda
> speakers, many of whom probably speak another language as well? I don't
> think the answer to that question is obvious - or useful, for that matter.
I wouldn't make statements on the English skills of either, really. I
wouldn't speculate on how English skills map to access to computers,
though. Which is another point, enabling people without "imperialistic"
language skills to feel comfortable in using the internet might empower
them more than the awesomebar.
Again, I really think that the values are good, but they're data points.
They don't really serve as a "oh, and the next on the list is the best
thing to do".
<...>
Axel
Simos Xenitellis
> The results of my language analysis for Firefox 3 are available:
> http://weblogs.mozillazine.org/gerv/archives/2008/06/firefox_3_language_coverage.html
>
Here are some comments for the Greece, Cyprus and the Greek language.
In the ODS file, the population of Greece is listed as 2,000,000 while
it is about 10,000,000.
A similar smaller population (about 350,000) is listed for Cyprus.
The data in the ODS file should be updated to the following:
Greece (population): 10,722,816 (CIA World Factbook)
Cyprus (population): 792,604 (CIA World Factbook)
Greek is also spoken in Cyprus, though a percentage is not shown
at either the CIA World Factbook, or Wikipedia.
My estimate is that the vast majority of the population speaks Greek or
are bilingual.
Currently there is no percentage for the Greek language in Cyprus, in
the ODS file.
Cheers,
Simos
Axel Hecht
I'm generally skeptical on recruiting localization teams. To me, users
benefit from a growing community, and they get hardly anything from a
one-time effort.
There is historical evidence that the recruiting we've done in the area
of localization hasn't really achieved growing communities, and they we
got stuck with one-time efforts, relatively short-lived ones, even.
So before we head out and make a priority list, we should make sure that
we're spending out limited resources on the right thing. The right thing
is to get users a sustained on-line life with mozilla, the wrong thing
is to make marks on the wall and just count languages.
Understanding how complex the creation and maintainance of a particular
localization should be is another point, as is making sure that people
actually find the localizations they're looking for.
We're testing the limits of our ecosystem at various ends, and we
shouldn't just test them, but expand them. But at the same time, "let's
have a binary for those" isn't always the answer to the question.
I'm feeling fine to call that "return of investment", not just from our
side, but from the side of the community as well.
Axel
Axel Hecht
We're not going to pay. As I said in another reply to this thread, it's
not about marks on the wall. As you said above, paid work isn't good
work, either, something we see across the ecosystem surprisingly often.
And we do have volunteers for Bengali, both India and Bangladesh.
There are a bunch of things to "invest" in, becoming more agile in how
we get localizations on board is probably the most important. I have
some ideas in that area, and I'm working on others. Yet others are
actually resolved, like not freezing everything for 1.9.0.x.
"Invest" in terms of time, thought, love, frust. Not money.
Axel
Ricardo Palomares Martínez
> We're not going to pay. As I said in another reply to this thread, it's
> not about marks on the wall. As you said above, paid work isn't good
> work, either, something we see across the ecosystem surprisingly often.
> "Invest" in terms of time, thought, love, frust. Not money.
Sorry if my post could be interpreted like I was suggesting to ask for
money. Mozilla and Mozilla Europe accounting books will show that I've
been not exactly eager to claim for fundings to travel to FOSDEM and
so. :-)
What I was (unsuccessfully) trying to expose is closer to what you
have said in another post, that "the next language in the list" can be
something difficult to achieve, because the right way to do it is
having volunteers and they should have appeared already by themselves.
If we really have volunteers localizing a language that will add
Firefox population and there is no official binary just because the
bug is waiting in the queue, then the process is where Mozilla needs
to act upon. But if there are registered localizers and a bug for a
language, but no langpack nor activity, then maybe we don't really be
so sure that we have volunteers for that language.
Ricardo
Gervase Markham
> Here are some comments for the Greece, Cyprus and the Greek language.
> In the ODS file, the population of Greece is listed as 2,000,000 while
> it is about 10,000,000.
Simos,
Thank you for your comments. The figures given are not population, but
number of Internet users. My source is here:
https://www.cia.gov/library/publications/the-world-factbook/rankorder/2153rank.html
It lists Greece as having 2,048,000 Internet users in 2006. If you have
a better or more recent figure, I would be happy to hear it.
> Greek is also spoken in Cyprus, though a percentage is not shown
> at either the CIA World Factbook, or Wikipedia.
> My estimate is that the vast majority of the population speaks Greek or
> are bilingual.
Do you mean the southern part of Cyprus, or the entire island? I don't
want to use the wrong words or offend anyone, but I was under the
impression that in part of Cyprus, Turkish is mostly spoken.
Is it possible to obtain population figures for the two parts of the
island? We could then divide up the internet population proportionally.
Gerv
Gervase Markham
> Robert Kaiser wrote:
>> can we reach?" or about "how many people do we support with software
>> in their own native language?"
>>
> The later question is a lot more interesting to me,
That's the question I'm trying to answer.
> and beyond that the
> question "how many more *could* we reach by the addition of each new
> incremental locale addition?" should get more attention.
And this one too. The spreadsheet gives absolute numbers of internet
population for each language, so we can see which language has the
highest number but no localization yet.
Gerv
Gervase Markham
> I'm generally skeptical on recruiting localization teams. To me, users
> benefit from a growing community, and they get hardly anything from a
> one-time effort.
>
> There is historical evidence that the recruiting we've done in the area
> of localization hasn't really achieved growing communities, and they we
> got stuck with one-time efforts, relatively short-lived ones, even.
>
> So before we head out and make a priority list, we should make sure that
> we're spending out limited resources on the right thing. The right thing
> is to get users a sustained on-line life with mozilla, the wrong thing
> is to make marks on the wall and just count languages.
Axel is right here. There is historical evidence that recruitment hasn't
worked well. Is that because recruitment is a bad idea full stop, or we
just did a bad job of it last time? (I did some of it, for Summer of
Code, so this is self-criticism here.)
But more information about the state of things can't be bad. How we use
that info is another question.
Gerv
Gervase Markham
> I don't follow your logic. There is no other OS which supports
> Dzongkha, so to use a Dzhongkha Mozilla browser, you have to use that
> Linux distribution.
That doesn't follow. If someone was struggling along with Windows in
their second or third language, they'd be very happy to use a browser in
their first language.
>> What percentage of the people of Bhutan speak Dzongkha as their first
>> language? (I can try and look this up, but if you have good information,
>> let me know.)
>
> Quickly, Wikipedia says first language of 130,000 and second language of
> ~470,000.
>
> http://en.wikipedia.org/wiki/Dzongkha_language
If we take that 130,000 (out of a population of 652,000) and apply it to
the Internet population of 30,000, we get a target market for Dzongkha
Firefox of 5,800 people.
Gerv
Gervase Markham
>> We're not going to pay. As I said in another reply to this thread, it's
>> not about marks on the wall. As you said above, paid work isn't good
>> work, either, something we see across the ecosystem surprisingly often.
>> (...)
>> "Invest" in terms of time, thought, love, frust. Not money.
>
>
> Sorry if my post could be interpreted like I was suggesting to ask for
> money.
Don't worry, I don't think Axel thought that. When he said "We're not
going to pay", he was making a general statement, not one directed at
you :-)
Gerv
Gervase Markham
> Mozilla should pay bengali translators just because nobody is
> interested to volunteer to do it? If so, how will all the not paid,
> volunteer translators to the current XX (45, 48, 76, whatever)
> languages feel about such decission?
Quite so.
As a general principle, introducing payment into a free software project
community is always a difficult thing. Apache, for example, avoids it
like the plague. Only recently did they agree to hire even a sysadmin,
as long as he did no coding.
Perl, on the other hand, has a history of making grants to key project
contributors to enable them to work on Perl full time for a short
period, to get key things done.
> Or maybe the real point is what can Mozilla do to attract people from
> those countries without Mozilla localizations to volunteer?
That's a good question. Or, to encourage and help the volunteers we do
have to get the work done. As Axel says, we have teams for quite a few
of the missing languages, but no localization yet.
Gerv