[Date Prev][Date Next] [Thread Prev][Thread Next] [Date Index] [Thread Index]

Debconf 8 internationalization sessions report

During Debconf 8, we handled four work sessions around
internationalization as well as more informal work around i18n

This long-overdue report summarizes the discussions during these
meetings as well as other i18n-related work that happened during
Debcamp and Debconf.

Please accept our apologies for us being very late sending this.

During Debconf 8, a lot of work happened around i18n, either during Debcamp,
informal sessions during Debconf and four formal 1-hour sessions.

i18n work session 1: infrastructure

During that work session, several contributors of Debian i18n
presented the recent progress about i18n infrastructure in Debian, 
mostly the dedicated server named "churro", aka i18n.debian.net.

churro is hosted in Merida, Extremadura, Spain, since September 2006, the
first i18n session held in Extremadura.

Since then, several services were developed on churro (see

Nicolas François summarized the work around l10n material collection.
Several scripts on churro gather the translation material and process it to

- access for translators to this material
- various statistics pages


Christian Perrier gave a summary about Pootle, the web interface
that allows access to l10n material as well as (potentially). Pootle is now
directly hooked to VCS repositories of Debian Installer as well as related
projects. Some teams began working with it to access the material. It still
needs very close monitoring, still, to manually handle possible conflicts.

Some experimental scripts are also currently running to gather all
po-debconf l10n material, commit it to debian-l10n SVN repository, with
Pootle hooked on it. Translators do not have access to this, though, because
we need to figure out how maintainers could easily get that material (for
instance by having debhelper scripts that could gather it at build time, or
similar solutions).

We now need to more closely define the administrators for Pootle, if it
becomes a production service (as it is partly, already, for D-I).


Michael Bramer gave the status of the DDTP (packages descriptions
translation project). Everything is now installed on churro since December
2007, including the web interface (DDTSS) developed by Martijn van

Future projects include the sync of translated descriptions with the archive
(see the detailed section about this later in this report) as well as
import/export of PO files, possibly hooked on Pootle (however, that leads to
severe performance issues).


Nicolas François showed the status of "tracking robots" that allow the most
organised l10n teams to track down their work with special-formatted mails
sent to their respective mailing lists. Recent weeks have seen Nicolas
merging together several scripts that were running by a few teams in a
general-purpose tracking robot that runs on churro


Christian Perrier showed recent evolutions on compendia creation. Compendia
are a collection of already translated strings, assembled in big PO files,
that can be used by translation teams to make the translation of new
material easier. The initial work by Eddy Petrisor is now running as a
general service on churro (http://i18n.debian.net/debian-l10n/compendia/)


Felipe Augusto van de Wiel summarized the status of the server
administration and hosting. The server is hosted in the facilities of the
Region of Extremadura. It is administered by Felipe, Christian Perrier,
César Gomez Martin (as local contact) and Nicolas François.

The server adminstration communication channel is the
debian-l10n-devel@lists.alioth.debian.org mailing list and #debian-i18n on
irc.debian.org. Nicolas and Felipe will setup IRC meetings between them
every 2 weeks.

A lot of work was done during Debcamp by Felipe and Nicolas to organize the
various scripts and robots in a cleaner way, to allow redundancy in server
and scripts monitoring.

The general goal is to have a "DSA-quality" server and progressively runs
all service in the most professional way. Philosophically speaking, we
target to transform i18n.debian.net into i18n.debian.org, ie an official
service of the project.

i18n work session 2: tdebs
(notes taken by Marc Hymers)

"tdebs" or "translation debs" are an attempt to achieve two goals, by
splitting localization material out of packahes:

- reduce space usage on systems with low resources (goal pushed from EmDebian)
- allow desynchronized updates of translations without requiring a full upload
  or package recompile (therefore potentially allowing l10n updates to stable,
  for instance)

Various constraints and desirable characteristics were discussed:
 - Out-of-band translation updates 
 - Only installing useful translations on systems 
 - Reducing bandwidth usage 
 - Avoiding archive bloat 

Discussion noted that "Reducing bandwidth usage" by providing small,
individual tdebs per package per architecture and "Avoiding archive
bloat" were contradictory. Instead it was suggested that one .tdeb per
source package would be provided, including all translations, and that
dpkg class support (http://www.dpkg.org/dpkg/Classes)
should be added to allow filtering at install time.

Using the class support, .tdebs would be easily splittable into
smaller, per language tdebs. This may be considered at stable

Translators would be able to use +t[0-9]+ syntax to indicate
translation updates (similar to bin-nmu style versioning)

Translation-Version: would be used in Packages files to indicate the
most up-to-date tdeb instead of having full Package: entries.

For new translations, we'll need some form of source; to be decided.

This work is aimed at lenny+2 due to the changes needed in the deb toolchain
(those being for lenny+1).

There needs to be discussion with the dpkg, apt, debian-installer and
debhelper people.

ftpmaster is happy with the one tdeb per source package and +t[0-9]+ syntax 

The plan is to meet in Extremedura (post-lenny) to continue the discussion
and start the work. The meeting should gather i18n representatives, Neil
Williams who did the first implementation of tdebs for EmDebian purposes,
ftpmasters and hopefully at least one dpkg maintainer (Raphaël Hertzog who
could be there for the QA meeting anyway?).

i18n work sessions 3: coordination, i18n team
The concept of an i18n "team" was discussed during that session.

 Language coordinators

First of all, we might want to establish the concept of team
"coordinators" (or "leaders"). The main point would then be to
officially establish a list of "official" leaders for each language
supported in Debian ("supported" meaning there exists a localization
effort for that language in one of the l10n areas that are specific to

Such "formal" list already exists for D-I translation work. It could
be used a a first draft list, even though some of the bigger teams
might want to change this.

Most participants in the meeting agreed that it would help getting in
touch with people who have a good knowledge in each language l10n
activities, as well as a good knowledge of that language itself.

The list of coordinators would then become public on a wiki page and
the nominated persons would then have the duty, if they accept, to
follow the activities of Debian i18n efforts (reading the debian-i18n
mailing list might be enough for this).

Some roles have been mentioned for language coordinators:
 * defining work methods for the team (for instance, that must be done
   for the DDTP before any DDTP work  starts)
 * act as a proxy between the translators and the core team
 * act as a point of contact for Debian project contributors
   who need to have input about a specific language
 * participate to i18n IRC meetings

 Core team

The concept of a "core team" for i18n has also been decided. This is
indeed formalizing something that already exists with the de facto lead taken by Christian Perrier on i18n and l10n activities in the project, as well as the work done by a few "key" people.

That core team would, for the beginning, include Christian Perrier
<bubulle@debian.org>, Felipe Augusto van de Wiel <faw@debian.org>,
Nicolas François <nicolas.francois@centraliens.net>, Michael Bramer

As of now, this is not a closed list but more formalizing what's
already happening.

Core team duties:
 * be responsible for the infrastructure and coordinate developments
 * organizing IRC meetings (reference: 
 * schedule regular public meetings of the core team
 * maintain the list of language coordinators (add new ones, ping
   existing ones, resolve disputes in case there is a conflict in an
   l10n team)

We need to find the right balance between a too formal process (which
might scare people from "one-person" teams) and a too loosy one where
nobody knows who's doing what.

i18n work sessions 4: DDTP
The DDTP/DDTSS infrastructure is now running solid on churro. Michael
re-summarized the various possibilities that translators have to adjust
their work process to the interface.

The major improvement is that translated descriptions *now flow into
the archive* since August 14th. Nicolas and Felipe worked hard on
scripts for this, along with ftpmasters.

So, from now, package management tools that support translated
descriptions (apt, aptitude, etc.) will display up-to-date versions
and not old stuff from 2006.

The discussion about the "openess" of the system is still strong. In our
i18n community, we clearly have two visions:
- "open to everybody with low constraints": this is the original design of
  the DDTP: everybody can show up and start translating. This allows for a
  quick growth of the amount of translations to the expense of their quality ;
- "control the quality": this is wished by the most structured teams
  and a concern bringed by many maintainers.

Felipe explained how he tries to balance this for the Brazilian Portuguese
team by basically controlling that nobody is translating outside the team
and following the work process for the team as well as assigning tasks in
the team.

The general agreement is that each team should be able to choose its
own process in the DDTP: some might want to restrict users by using
the built-in "authentication" system, while some other teams might
want to leave things more opened.

Another agreement is to converge towards the possible use of PO as a
method to translate in the DDTP (whether or not this is hooked to
Pootle is not necessarily relevant: we coould have a transient
situation where DDTP translation with PO files could be possible
without hooking Pootle to this). Michael will work on PO import (PO
export is alredy implemented and just needs to be automated).
Using PO would of course be optional.

i18n work sessions 4: Extremadura
An Extremadura meeting will happen in November (post Debconf: that
should be Nov 26th to Nov 30th).

This meeting will focus on tdeb work as well as DDTP/infrastructure
improvement. For that reason, attendees should be the "core i18n
infrastructure team", namely Felipe Augusto van de Wiel, Nicolas
François, Christian Perrier, Michael Bramer, Martijn van
Oosterhout. FTPmasters will be there as well, which will allow
polishing the import of packages description translations in the

The presence of Neil Williams and at least one dpkg developer (for
tdeb work) is wished.

The meeting would be shared with the QA meeting. We might have some
constraints in terms of number of participants, so we will need to
reduce it to the strict minimum. Steve McIntyre mentioned that he
cuold consider using Debian funds to sponsor some participation in
case that could be needed.

(post Debconf note: the meeting organization has started and is
coordinated on event-extremadura-meetings@teams.debian.net, which
anyone can subscribe to on http://teams.debian.net. The size
constraints are quite strong)

Achievements of Debcamp/Debconf
 * Thanks to the support of the ftpmasters, we could find a agreement on
   how to provide the translations of the package description in the
   archive. This was awaited for more than 6 years.

   This is still a temporary solution for Lenny, and should be
   changed to a solution were the translation are attached to packages to
   allow the transition of translation with the packages.
 * Several improvements were discussed before debconf for the coordination
   pages (http://i18n.debian.net/debian-l10n/) and the statistics pages on
   The full support for history could be finished in Debcamp and put in
   production for the coordination pages. A patch has been sent for review
   for the statistics pages.
   These improvements should help teams to focus on the translations where
   an effort is needed.
   More improvements are expected on the robots to match with the
   workflow of more teams in debian.
 * Cleanup of churro
   - some services were still using an old version of the tools based on
     the CVS version. They have been updated to the version publicly
     available on SVN.
 * Services on Churro are now provided with SSL support.
   The (self-signed) certificate used currently will be changed when we
   will have a certificate signed by DSA.
 * Improvements on DDTP (http://ddtp.debian.net)
 * Debcamp and Debconf permitted to exchange ideas and introduce the code
   base of Pootle and the dl10n robots.

The TODO list (http://wiki.debian.org/I18n/ToDoChurro) increased and
will keep us active for next year.

Meetings video recordings:

(to get high resolution versions, please s/low/high in URLs)


Attachment: signature.asc
Description: Digital signature

Reply to: