Re: Please don't list available translations in the package description

2007-12-18 Thread Ian Jackson
Leo \costela\ Antunes writes (Re: Please don't list available translations 
in the package description):
 While I agree it's an issue (albeit a small one), I think we shouldn't
 file bugs for it while there's no better place to put this information.
 It may reduce the objectiveness of some searches, but it is still
 valuable information.

I don't think this is a good argument.  Space in the package
description is not free: it costs download time (for Packages) files,
disk space, user attention (both when reading a specific description
and when searching) and so on.

Just because there is no better official place to put some information
does not mean that it should be in the Description:.  In this case I
dispute that this information, manually maintained, is valuable at all
let alone in the Description:.

I think Enrico should file bugs.

Ian.


-- 
To UNSUBSCRIBE, email to [EMAIL PROTECTED]
with a subject of unsubscribe. Trouble? Contact [EMAIL PROTECTED]



Re: Please don't list available translations in the package description

2007-12-08 Thread Lars Wirzenius
Bah, sorry about the previous mail, sent it before it was finished.

On pe, 2007-12-07 at 15:28 +0100, Thijs Kinkhorst wrote:
 On Friday 7 December 2007 11:52, Enrico Zini wrote:
  If we want to implement a feature such that we can see what packages are
  localised for a given language, that should be done differently, most
  likely by building a database scanning what .po files are present in the
  source package.
 
 Wouldn't it be possible to use a new debtags class like Localised-for:: to 
 store this information, and then use technique you describe to automatically 
 create those tags? The whole tag infrastructure already exists and seems very 
 suited to the kind of queries one would do on this information (e.g.: mail 
 client with console interface localised in Dutch and French).

Number of packages: 20 000
Number of languages we can expect to support in most packages: 10(*)
strlen(Localized-for: xx): 17
Overhead per Packages file: 3.2 MiB

Since the goal is, as far as I understand it, to include translations
for as many languages as possible, and almost all packages will include
translations for all of the major languages, repeating the same ones for
every package is very wasteful of space. Thus, I think it should not be
done. A better way to express the information is needed.

(*) English, French, German, Spanish (at least one variant), Portuguese,
Italian, Russian, Hindi, Japanese, Korean. In the long run, the list
will probably be quite a bit longer.



-- 
To UNSUBSCRIBE, email to [EMAIL PROTECTED]
with a subject of unsubscribe. Trouble? Contact [EMAIL PROTECTED]



Re: Please don't list available translations in the package description

2007-12-08 Thread Lars Wirzenius
On pe, 2007-12-07 at 15:28 +0100, Thijs Kinkhorst wrote:
 On Friday 7 December 2007 11:52, Enrico Zini wrote:
  If we want to implement a feature such that we can see what packages are
  localised for a given language, that should be done differently, most
  likely by building a database scanning what .po files are present in the
  source package.
 
 Wouldn't it be possible to use a new debtags class like Localised-for:: to 
 store this information, and then use technique you describe to automatically 
 create those tags? The whole tag infrastructure already exists and seems very 
 suited to the kind of queries one would do on this information (e.g.: mail 
 client with console interface localised in Dutch and French).

Number of packages: 20 000
Number of languages we can expect to support in most packages: 10(*)
strlen(Localized-for: xx): 17
Overhead per Packages file: 340 

(*) English, French, German, Spanish (at least one variant), Portuguese,
Italian, Russian, Hindi, 


-- 
To UNSUBSCRIBE, email to [EMAIL PROTECTED]
with a subject of unsubscribe. Trouble? Contact [EMAIL PROTECTED]



Re: Please don't list available translations in the package description

2007-12-08 Thread Thijs Kinkhorst
On Friday 7 December 2007 11:52, Enrico Zini wrote:
 If we want to implement a feature such that we can see what packages are
 localised for a given language, that should be done differently, most
 likely by building a database scanning what .po files are present in the
 source package.

Wouldn't it be possible to use a new debtags class like Localised-for:: to 
store this information, and then use technique you describe to automatically 
create those tags? The whole tag infrastructure already exists and seems very 
suited to the kind of queries one would do on this information (e.g.: mail 
client with console interface localised in Dutch and French).


Thijs


pgpX0oJhIxbmq.pgp
Description: PGP signature


Re: Please don't list available translations in the package description

2007-12-08 Thread Enrico Zini
On Sat, Dec 08, 2007 at 11:54:57AM +0200, Lars Wirzenius wrote:

 every package is very wasteful of space. Thus, I think it should not be
 done. A better way to express the information is needed.

Indeed.  If anyone can tell me how comparison among locale strings is
supposed to work, I can try to design something to integrate in
apt-xapian-index.


Ciao,

Enrico

-- 
GPG key: 1024D/797EBFAB 2000-12-05 Enrico Zini [EMAIL PROTECTED]


signature.asc
Description: Digital signature


Re: Please don't list available translations in the package description

2007-12-07 Thread Michael Banck
On Fri, Dec 07, 2007 at 10:52:08AM +, Enrico Zini wrote:
 ldap-account-manager is not the only package doing this: a quick search
 gives me [...] gperiodic

I removed that in svn.


Michael


-- 
To UNSUBSCRIBE, email to [EMAIL PROTECTED]
with a subject of unsubscribe. Trouble? Contact [EMAIL PROTECTED]



Re: Please don't list available translations in the package description

2007-12-07 Thread cobaco (aka Bart Cornelis)
On Friday 07 December 2007, Leo costela Antunes wrote:
 Enrico Zini wrote:
  I'm thinking of filing bugs, but I'd like to get some feedback here
  first.

 While I agree it's an issue (albeit a small one), I think we shouldn't
 file bugs for it while there's no better place to put this information.
 It may reduce the objectiveness of some searches, but it is still
 valuable information.

Almost no packages currently do this, hence relying on the package 
description to check wether a package is localized for your language is 
completely unreliable.

For a list of localised packages check http://www.debian.org/intl/l10n/po/ 
for your language, true it only cover's gettext localisations but that's 
99% of all localisation in free software anyway (this misses some localised 
webapps but that's about it AFAIK)

For a specific package you can also use apt-file or packages.debian.org to 
check for the presence of a lang-code.po file for your language.
-- 
Cheers, cobaco (aka Bart Cornelis)


signature.asc
Description: This is a digitally signed message part.


Re: Please don't list available translations in the package description

2007-12-07 Thread Francesco P. Lovergine
On Fri, Dec 07, 2007 at 01:40:43PM +, Stephen Gran wrote:
 This one time, at band camp, Leo costela Antunes said:
  Enrico Zini wrote:
   I'm thinking of filing bugs, but I'd like to get some feedback here
   first.
  
  While I agree it's an issue (albeit a small one), I think we shouldn't
  file bugs for it while there's no better place to put this information.
  It may reduce the objectiveness of some searches, but it is still
  valuable information.
 
 At the moment, I'm inclined to agree with Enrico.  I don't think it's
 all that helpful to have some small subset of all the programs actually
 translated into a given language returned in a search for that language
 string.  It's both an incomplete list (since many other programs will be
 localized, but just not mention it in the package description) and also
 useless clutter when you're looking for things related to actually
 working in the language.

I'd also tend to think that generic internationalization support
sentences are quite inappropriate into package long description.
Support leve is generally subject to frequent changes, so why
polluting descriptions with those things?

-- 
Francesco P. Lovergine


-- 
To UNSUBSCRIBE, email to [EMAIL PROTECTED]
with a subject of unsubscribe. Trouble? Contact [EMAIL PROTECTED]



Re: Please don't list available translations in the package description

2007-12-07 Thread Leo costela Antunes
cobaco (aka Bart Cornelis) wrote:
 Almost no packages currently do this, hence relying on the package 
 description to check wether a package is localized for your language is 
 completely unreliable.

Agreed.

 For a list of localised packages check http://www.debian.org/intl/l10n/po/ 
 for your language, true it only cover's gettext localisations but that's 
 99% of all localisation in free software anyway (this misses some localised 
 webapps but that's about it AFAIK)
 
 For a specific package you can also use apt-file or packages.debian.org to 
 check for the presence of a lang-code.po file for your language.

These are hardly practical solutions from a user perspective, even
though they are very helpful for developers.

What I meant is that I believe in the need for a way to tell users which
languages the package they intend to install supports (which doesn't
include greping for the package in a separate huge list of for files in
the source ;-) )

Cheers

-- 
Leo costela Antunes
[insert a witty retort here]



signature.asc
Description: OpenPGP digital signature


Re: Please don't list available translations in the package description

2007-12-07 Thread Stephen Gran
This one time, at band camp, Leo costela Antunes said:
 Enrico Zini wrote:
  I'm thinking of filing bugs, but I'd like to get some feedback here
  first.
 
 While I agree it's an issue (albeit a small one), I think we shouldn't
 file bugs for it while there's no better place to put this information.
 It may reduce the objectiveness of some searches, but it is still
 valuable information.

At the moment, I'm inclined to agree with Enrico.  I don't think it's
all that helpful to have some small subset of all the programs actually
translated into a given language returned in a search for that language
string.  It's both an incomplete list (since many other programs will be
localized, but just not mention it in the package description) and also
useless clutter when you're looking for things related to actually
working in the language.
-- 
 -
|   ,''`.Stephen Gran |
|  : :' :[EMAIL PROTECTED] |
|  `. `'Debian user, admin, and developer |
|`- http://www.debian.org |
 -


signature.asc
Description: Digital signature


Re: Please don't list available translations in the package description

2007-12-07 Thread Leo costela Antunes
Stephen Gran wrote:
 At the moment, I'm inclined to agree with Enrico.  I don't think it's
 all that helpful to have some small subset of all the programs actually
 translated into a given language returned in a search for that language
 string.  It's both an incomplete list (since many other programs will be
 localized, but just not mention it in the package description) and also
 useless clutter when you're looking for things related to actually
 working in the language.

Don't get me wrong, I agree with your assessment, I just think we could
come up with a better place to put this information _before_ filling
bugs. They may not be in the right place, but it's useful information
that IMHO should be accessible somewhere so that our users don't have to
go hunting for it on the web or are forced to install packages to check
it out.

Cheers

-- 
Leo costela Antunes
[insert a witty retort here]



signature.asc
Description: OpenPGP digital signature


Re: Please don't list available translations in the package description

2007-12-07 Thread Leo costela Antunes
Enrico Zini wrote:
 I'm thinking of filing bugs, but I'd like to get some feedback here
 first.

While I agree it's an issue (albeit a small one), I think we shouldn't
file bugs for it while there's no better place to put this information.
It may reduce the objectiveness of some searches, but it is still
valuable information.

Couldn't debtags support this sort of information in a good enough way
(not via culture::*, but something like translated-in::*)?

Or perhaps - if stricter solution is desired - implementing a new
control field that could be filled on build-time by a dh script (which
should support po files natively, but could also support regex based
searching of translation files, for alternative translation schemes)?

IMHO the debtags solution looks simpler and better, but it doesn't hurt
to keep our options open! :-)

Cheers
-- 
Leo costela Antunes
[insert a witty retort here]



signature.asc
Description: OpenPGP digital signature


Please don't list available translations in the package description

2007-12-07 Thread Enrico Zini
Hello,

I'm reviewing tag submissions and I noticed that people added lots of
culture::* tags to a non culture specific package like
ldap-account-manager.  The reason was simple: the description mentions:

  LAM is translated to Catalan, Chinese (Simplified + Traditional),
  Czech, Dutch, English, French, German, Hungarian, Italian, Japanese,
  Portuguese, Russian and Spanish.

This shows that people are lead into thinking that the package provides
special support for those cultures.  Read on for what I mean with
'special support'.

I don't think that this is a useful piece of information for a package
description: people doing apt-cache search japanese are likely not to
expect to see ldap-account-manager among the results: what one would
expect to see is dictionaries, input methods, festival modules, latex
styles and similar things.

If we want to implement a feature such that we can see what packages are
localised for a given language, that should be done differently, most
likely by building a database scanning what .po files are present in the
source package.

ldap-account-manager is not the only package doing this: a quick search
gives me sitemap, worker, easytag, gperiodic, ri-li and sailcut; there
are more, but unfortunately I can't provide a comprehensive list as it's
hard to spot these cases automatically.

I'm thinking of filing bugs, but I'd like to get some feedback here
first.


Ciao,

Enrico

-- 
GPG key: 1024D/797EBFAB 2000-12-05 Enrico Zini [EMAIL PROTECTED]


signature.asc
Description: Digital signature


Re: Please don't list available translations in the package description

2007-12-07 Thread Osamu Aoki
On Fri, Dec 07, 2007 at 03:10:35PM +0100, Francesco P. Lovergine wrote:
 On Fri, Dec 07, 2007 at 01:40:43PM +, Stephen Gran wrote:
  This one time, at band camp, Leo costela Antunes said:
   Enrico Zini wrote:
I'm thinking of filing bugs, but I'd like to get some feedback here
first.
   
   While I agree it's an issue (albeit a small one), I think we shouldn't
   file bugs for it while there's no better place to put this information.
   It may reduce the objectiveness of some searches, but it is still
   valuable information.
  
  At the moment, I'm inclined to agree with Enrico.  I don't think it's
  all that helpful to have some small subset of all the programs actually
  translated into a given language returned in a search for that language
  string.  It's both an incomplete list (since many other programs will be
  localized, but just not mention it in the package description) and also
  useless clutter when you're looking for things related to actually
  working in the language.
 
 I'd also tend to think that generic internationalization support
  ^^^ yes.
 sentences are quite inappropriate into package long description.
 Support leve is generally subject to frequent changes, so why
 polluting descriptions with those things?

I agree.  Since most of the packages understand UTF-8 these days for
i12n, there is no need to mention each l10n support.  

(In old days, jvim was different from vim to support Japanese ... I do
not know if it ever reached debian.org or stayed in debian.or.jp...
though)

Osamu


-- 
To UNSUBSCRIBE, email to [EMAIL PROTECTED]
with a subject of unsubscribe. Trouble? Contact [EMAIL PROTECTED]



Re: Please don't list available translations in the package description

2007-12-07 Thread Enrico Zini
On Fri, Dec 07, 2007 at 02:22:49PM +0100, Leo costela Antunes wrote:

 Couldn't debtags support this sort of information in a good enough way
 (not via culture::*, but something like translated-in::*)?
 Or perhaps - if stricter solution is desired - implementing a new
 control field that could be filled on build-time by a dh script (which
 should support po files natively, but could also support regex based
 searching of translation files, for alternative translation schemes)?

I don't think it's a good idea to use Debtags: it would immensely
clutter the Packages file, and the workflow for Debtags tags isn't ideal
for this.

Also, maybe with locales one would like to match more smartly than you
do with tags: maybe if I'm looking for it_CH I could be happy to find
another it_IT if it_CH is missing.

The feature is interesting, though.  I didn't know of
http://www.debian.org/intl/l10n/po/ posted by cobaco, and that could be
leveraged.  I'm thinking fetching info from
http://www.debian.org/intl/l10n/po/ and feeding them into
apt-xapian-index, for example.

If someone explains me precisely how locales are supposed to match, I
can work out a nice way of doing that.


Ciao,

Enrico

-- 
GPG key: 1024D/797EBFAB 2000-12-05 Enrico Zini [EMAIL PROTECTED]


signature.asc
Description: Digital signature


Re: Please don't list available translations in the package description

2007-12-07 Thread Christian Perrier
Quoting Francesco P. Lovergine ([EMAIL PROTECTED]):

 I'd also tend to think that generic internationalization support
 sentences are quite inappropriate into package long description.
 Support leve is generally subject to frequent changes, so why
 polluting descriptions with those things?


Sustained 200%.

For the record, when we come on such things in the Smith Project
(where we review debconf templates and, while we're at it, packages
descriptions), we propose maintainers to drop the relevant part and
use a more generic statement abou tthe software's i18n.

Indeed, any information that is likely to change over time should just
go away from packages' descriptions, as much as possible.

This also stands for statements like Package foo does this and
that and will probably do bar in the near future.




signature.asc
Description: Digital signature