Archive:L10N/Tools/Plans

Plans

Here are some detailed plans on what we can do in the localization landscape. Maybe "ideas" is a better word. Mostly ideas specific to tools, gathered in mailing lists and at FOSS.in when a bunch of translators and developers gathered around.

These aren't official plans of the FLP, just a bunch of useful notes and food for thought.

Correct what's not working (fix)

Priority 1

• Put Transifex into production at translate.fpo (100% complete)
• (./) Add missing modules & branches, do some final checks
• Announce (Fedora and lwn/foo)
• (./) Make sure statistics and every piece of the puzzle works

• Migrate users from elvis
• (./) Mass-email elvis users to create Fedora accounts (/L10N/Join page is ready)
• (./) IRC for help in registration process (sign CLA etc)

Priority 2

• Add email notifications
• For every commit, new module (or edit), notify related people
• Expose public parts as RSS too

• Enable all projects in Transifex
• Contact hosted.fpo-hosted developers to add their project too (eg. opyum)
• (./) Communicate with rest of elvis devs and urge them to move to Fedora infra

• Enable Tx on both app servers

• Add possibility to set flag with Tx (hold/release a package)

Increase efficiency (enhance)

Priority 1

• Code high-level views for maintenance
• Big table with registered modules, releases, branches
• Same table for Tx but also test write access for each module/branch
• SSH key overview/administration (admins)

• Need maintenance resources for DL and Tx
• DL: Keep in sync with upstream, convince them to do releases
• DL+Tx: Make sure all branches/modules are there, all the time. This applies for all registered super-projects (Fedora, RHEL, CentOS, OLPC)

• Move DL SQLite to mysql, have one app server to update the DB and the rest just their caches

• Configure DL to get notifications for each commit from each VCS (probably via Tx) and update LIVE the module's statistics, instead of running the cron job every a few hours.

• Separate Tx commit mechanism from web app to a separate process/service. (big project)
• Benefits: Increased security (apache no access to the SSH keys), enable an upstream project to actually do the commit/push, Tx only requests the commits (good for GNOME, KDE, etc)

• Increase verbosity in tools: Add more links and common info in both DL+Tx (eg. mention on the top of each page how to checkout and where from to checkin respectively, links to bugzilla and maintainers, etc).

Priority 2

• Give more control to developers directly
• Add web-based interface for developers to register their projects
• Give them the ability to add a new branch to their module
• Give Transifex access directly through the FAS (reduced overhead on Fedora Infra)

• Expose repos: Give developers the ability to pull translations instead of having Tx push them

• Build RPM for DL and push DL+Tx RPMs in Fedora Universe

• Build a common model/configuration with DL (maintenance cost down).
• Create similar models to DL in Tx. Experiment in running DL scripts to populate them. Goal: Not having to maintain two separate Views, configuration files, checked-out caches.

Add functionality (extend)

Priority 1

• Implement VCS-agnostic CLI client to work independantly of the web interface (big project)
• Eg. tx checkout --all and tx submit <proj> <branch> <file>
• Somewhat related with Pootle integration
• Could get integrated with kbabel, gtranslator, etc.

• Pootle: Make it work with Transifex
• See what is/isn't needed and customize accordingly
• Install a test instance, even without Tx for specspo

• RH Docs
• Get in touch with folks to see if they could work closer with the community and leverage its throughput
• Why not have the docs in some "unofficially supported" languages too?

• Discuss with Debian community their needs and requirements

• Test OpenID

• Make Tx *completely* portable to any independent project that wants its resources localized

Priority 2

• Fine-grain permissions in Tx: Language maintainers, approvals by proj/lang/branch, etc. Who owns what, who controls what.
• Add ability to "hold" and "release" a project/branch, like in elvis

• Start thinking big
• Have language-specific and project-specific sub-domains, specific content for each

Community

• Continue building groups, mailing lists, the community
• Work better together with Ambassadors -- look at Ubuntu for "loco teams"
• Send emails to -announce before every release
• Give credits in all Fedora-is-upstream applications/docs
• Bi-weekly meetings
• Split specspo in chunks of reduced priority.
• Also, make it work with rpm (spot)

Other ideas worth considering

• JBOSS Docs, similar to RH Docs
• Docs in general! LTSP? Manpages?