Firefox OS language packs

classic Classic list List threaded Threaded
46 messages Options
123
Reply | Threaded
Open this post in threaded view
|

Firefox OS language packs

Gervase Markham
[CC people: discussion is in mozilla.dev.l10n]

It was suggested in mozilla.governance that Mozilla should make "ship a
certain number (or all) language packs" a condition of the Firefox OS
logo license program, which is our main way of influencing what happens
with Firefox OS.

I raised this with the business people, who said:

"An OEM's reaction to this ask will boil down what does it cost to
support/ship all the language packs:

* memory
* disk space
* QA of the languages to the OEM's satisfaction
* further localization of OEM/operator customizations, etc.
* What happens if a langpack is missing a string

Is anyone able to provide information on these matters, and anything
else which would influence their decision?

We aren't the only mobile OS in town; the requirements we set on the
logo program do, in part, affect an OEM's decision as to whether to go
with us or with someone else. So we have to be sure it's worth making
this a requirement.

(Or it may be that we are soon going to have user-installable langpacks
with seamless fallback, in which case this question becomes very different.)

Gerv
_______________________________________________
dev-l10n mailing list
[hidden email]
https://lists.mozilla.org/listinfo/dev-l10n
Reply | Threaded
Open this post in threaded view
|

Re: Firefox OS language packs

Michael Bauer-11
My tuppence...

20/02/2013 19:51, sgrìobh Gervase Markham:
> "An OEM's reaction to this ask will boil down what does it cost to
> support/ship all the language packs:
>
> * memory
> * disk space
Well, if we have the user-installable packs you mention below, this
won't be an issue. In terms of my own locale, I honestly don't care
whether it comes bundled or via a pack, as long as it's not up to the
OEM to include or exclude them. For the simple reason that on the whole,
they just don't think about and/or care about languages which aren't
economically relevant to them - which means most languages under ~5
million speakers.
> * QA of the languages to the OEM's satisfaction
I don't think the OEM's are actually that smart about the languages they
DO handle... a lot of them are arcane but I know that's not what you
meant ;)

Proofreading doesn't cost the world. How about a process where for every
language that gets admitted Mozilla sends a dozen random screenshots to
an independent proofreader for feedback?
> * further localization of OEM/operator customizations, etc.
> * What happens if a langpack is missing a string
Use a fallback language for each locale?
> We aren't the only mobile OS in town; the requirements we set on the
> logo program do, in part, affect an OEM's decision as to whether to go
> with us or with someone else. So we have to be sure it's worth making
> this a requirement.
Isn't it more a question about Mozilla's philosophy?
> (Or it may be that we are soon going to have user-installable langpacks
> with seamless fallback, in which case this question becomes very different.)
Which, as I said, would be fine with me too.

Michael
_______________________________________________
dev-l10n mailing list
[hidden email]
https://lists.mozilla.org/listinfo/dev-l10n
Reply | Threaded
Open this post in threaded view
|

Re: Firefox OS language packs

Axel Hecht
In reply to this post by Gervase Markham
Hi,

I'll answer the questions in details below, but I'd like to re-iterate
the current technical constraints:

- we can't change the localizations of an app at runtime
- we can't dynamically load additional data

... without breaking our startup performance, that is.

Fixing either is an to-be-scoped engineering task, with unknown outcome.

I'll answer below based on the status-quo of what we consider necessary
to perform.

On 20.02.13 20:51, Gervase Markham wrote:

> [CC people: discussion is in mozilla.dev.l10n]
>
> It was suggested in mozilla.governance that Mozilla should make "ship a
> certain number (or all) language packs" a condition of the Firefox OS
> logo license program, which is our main way of influencing what happens
> with Firefox OS.
>
> I raised this with the business people, who said:
>
> "An OEM's reaction to this ask will boil down what does it cost to
> support/ship all the language packs:
>
> * memory

Linear in terms of locales, as all of them are loaded into memory.

> * disk space

Linear in terms of locales. We're trying to get the factor down by not
needing all of toolkit, see the thread on .platform.

> * QA of the languages to the OEM's satisfaction

I doubt we'll actually see this among the OEMs. If they don't care about
a locale, they can just as well rely on our QA.

QA on our side is a hefty problem, though. Devices, follow-up, etc.

> * further localization of OEM/operator customizations, etc.

Right now, we're not falling back great. Say, we're falling back from
fy-NL to en-US instead of nl. This is pretty ugly.

> * What happens if a langpack is missing a string

Fallback to en-US.

> Is anyone able to provide information on these matters, and anything
> else which would influence their decision?
>
> We aren't the only mobile OS in town; the requirements we set on the
> logo program do, in part, affect an OEM's decision as to whether to go
> with us or with someone else. So we have to be sure it's worth making
> this a requirement.
>
> (Or it may be that we are soon going to have user-installable langpacks
> with seamless fallback, in which case this question becomes very different.)
>
> Gerv


One other issue, the language lists of OEMs are not based on hard
opinions. They're usually picked up data that some other company did
based on absolutely nothing.

We can also provide real data that shows why we think particular
languages are important to large populations in the target markets.

Axel

_______________________________________________
dev-l10n mailing list
[hidden email]
https://lists.mozilla.org/listinfo/dev-l10n
Reply | Threaded
Open this post in threaded view
|

Re: Firefox OS language packs

Michael Bauer-11

20/02/2013 22:32, sgrìobh Axel Hecht:
>
> ... without breaking our startup performance, that is.
Would that be a hit on startup perfomance once or every time?
>
> We can also provide real data that shows why we think particular
> languages are important to large populations in the target markets.

Well that's our locale doomed, along with Welsh, Asturian, Basque,
Breton, Frisian, Icelandic, Kashubian, Rumantsch etc etc...

Michael
_______________________________________________
dev-l10n mailing list
[hidden email]
https://lists.mozilla.org/listinfo/dev-l10n
Reply | Threaded
Open this post in threaded view
|

Re: Firefox OS language packs

Axel Hecht
In reply to this post by Axel Hecht
On 20.02.13 23:49, Michael Bauer wrote:
>
> 20/02/2013 22:32, sgrìobh Axel Hecht:
>>
>> ... without breaking our startup performance, that is.
> Would that be a hit on startup perfomance once or every time?

Every time.

>>
>> We can also provide real data that shows why we think particular
>> languages are important to large populations in the target markets.
>
> Well that's our locale doomed, along with Welsh, Asturian, Basque,
> Breton, Frisian, Icelandic, Kashubian, Rumantsch etc etc...

That's not doomed, that's just different aspects of the discussion.

Axel

_______________________________________________
dev-l10n mailing list
[hidden email]
https://lists.mozilla.org/listinfo/dev-l10n
Reply | Threaded
Open this post in threaded view
|

Re: Firefox OS language packs

Kai Hendry
On 21 February 2013 07:16, Axel Hecht <[hidden email]> wrote:
>> Would that be a hit on startup perfomance once or every time?

At Webconverger we solved the "all the langpacks killing performance"
problem, by linking in the langpack based on locale which is a system
wide setting:
https://github.com/Webconverger/webc/blob/master/etc/webc/live-config.sh#L81

Tbh I find FF langpacks a PITA. I wish it was integrated somehow. The
changelog between updates is bonkers:
https://github.com/Webconverger/webc/commit/7924f13ff275af10ac8f480ef80c979212a7769d

Good luck! :)
_______________________________________________
dev-l10n mailing list
[hidden email]
https://lists.mozilla.org/listinfo/dev-l10n
Reply | Threaded
Open this post in threaded view
|

Re: Firefox OS language packs

Francesco Lodolo [:flod]
2013/2/21 Kai Hendry <[hidden email]>

> At Webconverger we solved the "all the langpacks killing performance"
> problem, by linking in the langpack based on locale which is a system
> wide setting:
>
> https://github.com/Webconverger/webc/blob/master/etc/webc/live-config.sh#L81
>

I believe you're talking about Firefox (desktop), while this discussion is
about Firefox OS, a completely different beast..

Francesco
_______________________________________________
dev-l10n mailing list
[hidden email]
https://lists.mozilla.org/listinfo/dev-l10n
Reply | Threaded
Open this post in threaded view
|

Re: Firefox OS language packs

Gervase Markham
In reply to this post by Axel Hecht
Hi Axel,

This is really helpful - thanks.

On 20/02/13 22:32, Axel Hecht wrote:
>> * memory
>
> Linear in terms of locales, as all of them are loaded into memory.

So if I have a 10-locale phone, more memory is used, even if I'm only
ever using 1 of those locales?

Even if we decide to have a system which means I have to reboot the
phone to change locale?

What's the memory footprint per locale, roughly?

>> * further localization of OEM/operator customizations, etc.
>
> Right now, we're not falling back great. Say, we're falling back from
> fy-NL to en-US instead of nl. This is pretty ugly.

Are there plans to fix that, or is it another Hard Problem?

> We can also provide real data that shows why we think particular
> languages are important to large populations in the target markets.

Now that _is_ a good idea. We can, at least, provide a "list of
suggested languages" for each locale our partners are shipping in.

Gerv


_______________________________________________
dev-l10n mailing list
[hidden email]
https://lists.mozilla.org/listinfo/dev-l10n
Reply | Threaded
Open this post in threaded view
|

Re: Firefox OS language packs

Axel Hecht
On 21.02.13 12:45, Gervase Markham wrote:

> Hi Axel,
>
> This is really helpful - thanks.
>
> On 20/02/13 22:32, Axel Hecht wrote:
>>> * memory
>>
>> Linear in terms of locales, as all of them are loaded into memory.
>
> So if I have a 10-locale phone, more memory is used, even if I'm only
> ever using 1 of those locales?

Yes.

> Even if we decide to have a system which means I have to reboot the
> phone to change locale?

With the system we have, reboot or not doesn't matter. I'd rather keep
wishlists for other systems to other threads.

> What's the memory footprint per locale, roughly?

I don't know, but some data point to start from: 10k is the strings for
the system app as source, localizations are 14-17k. Settings app has 22k
for en-US. For the current build, the index.html of the systems app
grows from <40k to >450K.

All those are parsed in js data structures right now, no idea how that
maps out.

>>> * further localization of OEM/operator customizations, etc.
>>
>> Right now, we're not falling back great. Say, we're falling back from
>> fy-NL to en-US instead of nl. This is pretty ugly.
>
> Are there plans to fix that, or is it another Hard Problem?

The problem is that l10n.js is hooked on navigator.language, which is
only a single value. The best upcoming multi-locale API is the JS i18n
API that Norbert is working on. He wants to redo his ETA of April next
week. That's with ICU and all that. When we'd be willing to update
Firefox OS to a gecko that comes with that, no idea.

Axel

>> We can also provide real data that shows why we think particular
>> languages are important to large populations in the target markets.
>
> Now that _is_ a good idea. We can, at least, provide a "list of
> suggested languages" for each locale our partners are shipping in.
>
> Gerv
>
>

_______________________________________________
dev-l10n mailing list
[hidden email]
https://lists.mozilla.org/listinfo/dev-l10n
Reply | Threaded
Open this post in threaded view
|

Re: Firefox OS language packs

Toni Hermoso Pulido-2
Hi,

I cannot add any further value to the discussion, but I just want to
comment that I think can confirm the issue of more memory consumption
in Firefox OS when more locales are added. I tried including all
possible ones and there was an important degradation of performance in
an unagi device :/
I thought at that moment that it was a buid puntual thing but now I
see it was very likely because of having added all locales at that
time.


2013/2/21 Axel Hecht <[hidden email]>:

> On 21.02.13 12:45, Gervase Markham wrote:
>>
>> Hi Axel,
>>
>> This is really helpful - thanks.
>>
>> On 20/02/13 22:32, Axel Hecht wrote:
>>>>
>>>> * memory
>>>
>>>
>>> Linear in terms of locales, as all of them are loaded into memory.
>>
>>
>> So if I have a 10-locale phone, more memory is used, even if I'm only
>> ever using 1 of those locales?
>
>
> Yes.
>
>
>> Even if we decide to have a system which means I have to reboot the
>> phone to change locale?
>
>
> With the system we have, reboot or not doesn't matter. I'd rather keep
> wishlists for other systems to other threads.
>
>
>> What's the memory footprint per locale, roughly?
>
>
> I don't know, but some data point to start from: 10k is the strings for the
> system app as source, localizations are 14-17k. Settings app has 22k for
> en-US. For the current build, the index.html of the systems app grows from
> <40k to >450K.
>
> All those are parsed in js data structures right now, no idea how that maps
> out.
>
>
>>>> * further localization of OEM/operator customizations, etc.
>>>
>>>
>>> Right now, we're not falling back great. Say, we're falling back from
>>> fy-NL to en-US instead of nl. This is pretty ugly.
>>
>>
>> Are there plans to fix that, or is it another Hard Problem?
>
>
> The problem is that l10n.js is hooked on navigator.language, which is only a
> single value. The best upcoming multi-locale API is the JS i18n API that
> Norbert is working on. He wants to redo his ETA of April next week. That's
> with ICU and all that. When we'd be willing to update Firefox OS to a gecko
> that comes with that, no idea.
>
> Axel
>
>
>>> We can also provide real data that shows why we think particular
>>> languages are important to large populations in the target markets.
>>
>>
>> Now that _is_ a good idea. We can, at least, provide a "list of
>> suggested languages" for each locale our partners are shipping in.
>>


--
Toni Hermoso Pulido
http://www.cau.cat
_______________________________________________
dev-l10n mailing list
[hidden email]
https://lists.mozilla.org/listinfo/dev-l10n
Reply | Threaded
Open this post in threaded view
|

Re: Firefox OS language packs

Michael Bauer-11
In reply to this post by Axel Hecht

20/02/2013 23:16, sgrìobh Axel Hecht:
> That's not doomed, that's just different aspects of the discussion.
>
> Axel
Given how well l10n and Mozilla Mobile are doing, you'll have to pardon
me for not sharing your confidence though of course I'd love to be wrong
on this one.

21/02/2013 11:45, sgrìobh Gervase Markham:
> Now that_is_  a good idea. We can, at least, provide a "list of
> suggested languages" for each locale our partners are shipping in.
>
> Gerv
I'd just like to remind people of the reality of what commercial
companies' attitude to non-economically relevant languages is. Anyone
remember the trouble Icelandic had with Microsoft
(http://chronicle.augusta.com/stories/1998/06/30/tec_232220.shtml) or
that the Basque Government initially *paid* Microsoft to do a Basque
version
(http://paperekoa.berria.info/plaza/2011-02-02/037/002/ordenagailua_euskaratzen.htm)?

Having a bit of a windmill feeling right now.

Michael
_______________________________________________
dev-l10n mailing list
[hidden email]
https://lists.mozilla.org/listinfo/dev-l10n
Reply | Threaded
Open this post in threaded view
|

Re: Firefox OS language packs

Guillermo López Leal
In reply to this post by Toni Hermoso Pulido-2
2013/2/21 Toni Hermoso Pulido <[hidden email]>

> I tried including all
> possible ones and there was an important degradation of performance in
> an unagi device :/
>

IIRC, now if you flash gaia and add a lot of locales, each localization is
put on the bottom of the index.html, as a script with the type text/l10n.
Before, it was a XHR to a json resource outside the index.html. I don't
know how the parse works, but maybe it need to load everything on memory,
including those localizations.


--
Guillermo López [willyaranda]
http://mozilla-hispano.org
http://twitter.com/mozilla_hispano
http://facebook.com/mozillahispano
Certified Mozillian: https://mozillians.org/willyaranda
_______________________________________________
dev-l10n mailing list
[hidden email]
https://lists.mozilla.org/listinfo/dev-l10n
Reply | Threaded
Open this post in threaded view
|

Re: Firefox OS language packs

Gervase Markham
In reply to this post by Axel Hecht
On 21/02/13 12:59, Axel Hecht wrote:
>> So if I have a 10-locale phone, more memory is used, even if I'm only
>> ever using 1 of those locales?
>
> Yes.

Given that we are searching for max performance on memory-constrained
devices, that seems like a showstopper, right there.

> I don't know, but some data point to start from: 10k is the strings for
> the system app as source, localizations are 14-17k. Settings app has 22k
> for en-US. For the current build, the index.html of the systems app
> grows from <40k to >450K.

That's going to have a significant effect on page (app) load time.

Gerv

_______________________________________________
dev-l10n mailing list
[hidden email]
https://lists.mozilla.org/listinfo/dev-l10n
Reply | Threaded
Open this post in threaded view
|

Re: Firefox OS language packs

Robert Kaiser
In reply to this post by Axel Hecht
Axel Hecht schrieb:
> The problem is that l10n.js is hooked on navigator.language, which is
> only a single value. The best upcoming multi-locale API is the JS i18n
> API that Norbert is working on. He wants to redo his ETA of April next
> week. That's with ICU and all that. When we'd be willing to update
> Firefox OS to a gecko that comes with that, no idea.

I doubt we'd want that any time soon, esp. with the current line of
devices, as ICU is huge and would surely take away a ton more memory
than a few localizations.

Robert Kaiser
_______________________________________________
dev-l10n mailing list
[hidden email]
https://lists.mozilla.org/listinfo/dev-l10n
Reply | Threaded
Open this post in threaded view
|

Re: Firefox OS language packs

Gervase Markham
On 25/02/13 01:15, Robert Kaiser wrote:
> I doubt we'd want that any time soon, esp. with the current line of
> devices, as ICU is huge and would surely take away a ton more memory
> than a few localizations.

As I recall, that discussion ended with Norbert going away to have
another think and return with an alternative proposal. So we'll see what
he comes up with.

Gerv


_______________________________________________
dev-l10n mailing list
[hidden email]
https://lists.mozilla.org/listinfo/dev-l10n
Reply | Threaded
Open this post in threaded view
|

Re: Firefox OS language packs

Fjoerfoks
In reply to this post by Gervase Markham
Sounds like a good reason to let users uninstall unnecessary languages to
speed up performance,
and install only the languages they want. Indeed, the fallback system is an
issue to be investigated.
Please don't put it into the hands of providers, but let the user control
it.
I know, I know, I'm only preaching for my own church................

Wim

2013/2/22 Gervase Markham <[hidden email]>

> On 21/02/13 12:59, Axel Hecht wrote:
> >> So if I have a 10-locale phone, more memory is used, even if I'm only
> >> ever using 1 of those locales?
> >
> > Yes.
>
> Given that we are searching for max performance on memory-constrained
> devices, that seems like a showstopper, right there.
>
> > I don't know, but some data point to start from: 10k is the strings for
> > the system app as source, localizations are 14-17k. Settings app has 22k
> > for en-US. For the current build, the index.html of the systems app
> > grows from <40k to >450K.
>
> That's going to have a significant effect on page (app) load time.
>
> Gerv
>
_______________________________________________
dev-l10n mailing list
[hidden email]
https://lists.mozilla.org/listinfo/dev-l10n
Reply | Threaded
Open this post in threaded view
|

Re: Firefox OS language packs

Michael Bauer-11
There are quite a few in your church Wim :)

Michael

25/02/2013 12:54, sgrìobh Fjoerfoks:
> I know, I know, I'm only preaching for my own church................
>
> Wim

--
*Akerbeltz <http://www.faclair.com/>*
Goireasan Gàidhlig air an lìon
Fòn: +44-141-946 4437
Facs: +44-141-945 2701

*Tha Gàidhlig aig a' choimpiutair agad, siuthad, feuch e!*
Iomadh rud eadar prògraman oifis, brabhsairean, predictive texting,
geamannan is mòran a bharrachd. Tadhail oirnn aig www.iGàidhlig.net
<http://www.iGaidhlig.net/>
_______________________________________________
dev-l10n mailing list
[hidden email]
https://lists.mozilla.org/listinfo/dev-l10n
Reply | Threaded
Open this post in threaded view
|

Re: Firefox OS language packs

Julen Ruiz Aizpuru
In reply to this post by Gervase Markham
Hi,

Sorry if I chime in a little bit late.

og., 2013.eko otsren 21a 12:45(e)an, Gervase Markham(e)k idatzi zuen:
>
>> We can also provide real data that shows why we think particular
>> languages are important to large populations in the target markets.
>
> Now that _is_ a good idea. We can, at least, provide a "list of
> suggested languages" for each locale our partners are shipping in.
>

I understand there are technical problems that make it impossible to
ship/offer all the locales in Firefox OS at the moment, but it'll be
unfortunate if some locales are prioritized based on population — our
users won't understand that kind of reasoning coming from Mozilla.
(Any bug numbers for this issue, by the way? or any hints to offer help
fixing the related issues?)

Now, after reading last Sunday the announcement that Firefox OS-enabled
devices will be available in Spain starting in June, I'm really worried
that this would practically mean —at least in a beginning, if large
population measures are mainly taken into account and unless the
blockers to offer any localization to users are fixed— the extinction
from Firefox OS devices for us the small languages spoken in Spain.

So I don't know what the final outcome will be, but sooner or later I'm
sure our users will start asking us if they'll be able to buy and use a
Firefox OS device in their mother tongue. Do we have any answers for that?

It's 2013 and we still have no smartphones officially available in our
own language, so Firefox OS is probably the only choice in the short/mid
term — which, by the way, can be another selling point.

Julen.
_______________________________________________
dev-l10n mailing list
[hidden email]
https://lists.mozilla.org/listinfo/dev-l10n
Reply | Threaded
Open this post in threaded view
|

Re: Firefox OS language packs

Rhoslyn Prys
In reply to this post by Gervase Markham
Welsh is the third most popular language in the UK after English and
Polish. My Tesco HTC Wildfire includes a choice of German, English,
Spanish, French, Italian and Polish but no Welsh. Also all of the
languages have a different language pack for the UK and Ireland so
that's 12 packs in all. Orange introduced a Samsung GT-S5600 with a
Welsh interface a few years ago that was quite popular with Welsh
speakers even though as a phone it's not great. I think there was also
an Irish version as well.

It all comes down to Mozilla's wish/ability to provide users with an
interface in their language of choice. If there is a translation of the
interface available why not recommend it to carriers and users? Mozilla
has a long history of doing this. I don't see any reason for a change in
policy.

Rhos.
_______________________________________________
dev-l10n mailing list
[hidden email]
https://lists.mozilla.org/listinfo/dev-l10n
Reply | Threaded
Open this post in threaded view
|

Re: Firefox OS language packs

Anton Meixome
In reply to this post by Julen Ruiz Aizpuru
+1


2013/2/27 Julen Ruiz Aizpuru <[hidden email]>

> Hi,
>
> Sorry if I chime in a little bit late.
>
> og., 2013.eko otsren 21a 12:45(e)an, Gervase Markham(e)k idatzi zuen:
>
>
>>  We can also provide real data that shows why we think particular
>>> languages are important to large populations in the target markets.
>>>
>>
>> Now that _is_ a good idea. We can, at least, provide a "list of
>> suggested languages" for each locale our partners are shipping in.
>>
>>
> I understand there are technical problems that make it impossible to
> ship/offer all the locales in Firefox OS at the moment, but it'll be
> unfortunate if some locales are prioritized based on population — our users
> won't understand that kind of reasoning coming from Mozilla.
> (Any bug numbers for this issue, by the way? or any hints to offer help
> fixing the related issues?)
>
> Now, after reading last Sunday the announcement that Firefox OS-enabled
> devices will be available in Spain starting in June, I'm really worried
> that this would practically mean —at least in a beginning, if large
> population measures are mainly taken into account and unless the blockers
> to offer any localization to users are fixed— the extinction from Firefox
> OS devices for us the small languages spoken in Spain.
>
> So I don't know what the final outcome will be, but sooner or later I'm
> sure our users will start asking us if they'll be able to buy and use a
> Firefox OS device in their mother tongue. Do we have any answers for that?
>
> It's 2013 and we still have no smartphones officially available in our own
> language, so Firefox OS is probably the only choice in the short/mid term —
> which, by the way, can be another selling point.
>
> Julen.
>
> ______________________________**_________________
> dev-l10n mailing list
> [hidden email]
> https://lists.mozilla.org/**listinfo/dev-l10n<https://lists.mozilla.org/listinfo/dev-l10n>
>



--
Antón Méixome - Galician Native Lang Coordination
Galician community LibO & AOO
_______________________________________________
dev-l10n mailing list
[hidden email]
https://lists.mozilla.org/listinfo/dev-l10n
123