#268 Long term funding request: translation platform hosting
Closed: approved 2 years ago by bcotton. Opened 2 years ago by jibecfed.

The localization community is in a critical situation. We are less and less active contributors and our translation platform is dying.
Migrating to another translation platform is mandatory, both to revive our community and lower the risk of being without tool.

The migration plan, describing the current situation, some of our key requirements and the steps was validated during Flock. The only viable possibility is Weblate.
See the wiki pages: https://fedoraproject.org/wiki/L10N_Move_to_Weblate

As the Community Platform Engineering decided to focus on key application, they won't assume the hosting this platform.
Using the OpenShift community platform would be a short term solution, but while I have the skills to host it in the context of a Proof of Concept (you'll see our future documentation on https://translate.holcroft.fr), I'm not skilled enough to assume the responsibility for our community and not for a long term period.

I'm willing to assume, with the help of other l10n community members, the administration of the tool, but not the software+system maintenance actions.

The only solution we could find during Flock is to pay for the hosting of this platform.

But paying for the hosting only is not enough. While Weblate has a lot of interesting features that will hugely help our community, we need a lot of tiny changes to make the tool fitted for a big project such as the Fedora Project.

In the past, we lost Transifex who turned closed source. We lost Zanata because it wasn't a sustainable open-source project, the risk exists that Weblate also becomes closed source product in the future. The only serious alternative we know about, Pootle, haven't seen any commit in a year and a half.
I think this is a risky situation for the whole open-source ecosystem. And this is a risk for our project, because changing our translation platform breaks existing processes and contributor efficiency.

The LibreOffice community is facing the same challenges (they are now using Pootle).
The OpenSuse community is already using Weblate.

Therefore, the Fedora community should probably commit to reducing this risk for the coming years:

  • to animate a cross projects discussion on our needs and requirements,
  • to dedicate funds to have premium support and new features,
  • to go further than the translation platform and to include other key peripheral tools (translation memory server...),
  • make sure that every code is be open-source and uptreamed, from core Weblate features to side tools and libraries

Cost for hosting and support is 6000 € per year.
The amount of the dedicated funds for premium support and new features is to be decided with other communities. I only ask here if the Fedora Council acknowledge the situation, and is willing to open a cross project discussion.

For your information on hosting plans: https://weblate.org/en/hosting/

About our current translation platform:

  • 346 608 source messages
  • 4 232 121 source words.
  • 106 projects (you can expect as much VCS)

Theses statistics doesn't include the translation of our documentation: https://translate.holcroft.fr/stats/

Thanks a lot for your help
I could not explain everything in this already long ticket, please let me know additional required information.


+1 to opening the dialog. @jibecfed would you drive the conversation?

I'd also like to hear if we can get our infrastructure team to support the recurring costs are part of providing infrastructure so that our community funds can stay focused on community activities, not infra.

Thanks, I should be able to attend, I'll keep you and Ben upy to date on telegram.

This is related to the Docs translations we've prototyped with @jibecfed recently. Preview is in staging: http://docs.stg.fedoraproject.org

@jibecfed two questions:

  1. Where does the 6000 €/year come from? Is that a custom quote they gave us based on our size? If so, what does it include? The pricing page doesn't have any listing that matches that number.

  2. This is a lot of money. While I definitely support providing a platform for the l10n community, I don't want to commit us to this amount indefinitely. Do you have a plan for evaluating the success of this platform? In other words: what would should we use to decide if we want to continue funding this in future years?

@bcotton the answers to your question:

  1. The amount comes from Weblate's support I contacted with the number I shared here to have an estimation (ticket number #598). I don't have a detailed list of activities, but it includes bugfixing, upgrades and dedicated support.
  2. Thanks to Brian who opened doors inside Red Hat teams, I received a database dump and have a bunch of statistics about our translation platform. I wrote about there: https://communityblog.fedoraproject.org/fedora-localization-project-status-and-horizons/ Measuring how successful we are shouldn't be too difficult (we could even ask weblate to produce these graphs, only the categorization looks missing to me). Fallback scenario will always be to self-host it.

Is the 6000 €/year purely for the hardware/VMs or do they manage the entire application stack that is required for translations?

Is the 6000 €/year purely for the hardware/VMs or do they manage the entire application stack that is required for translations?

They do manage the entire application stack. To my knowledge: this is a postgresql+redis+django+celery for weblate. Plus a translation memory server which I think is fully included in weblate's code. They will handle some configuration of Weblate itself, but I think our community should be quite autonomous in administrating Weblate itself.

Note: I would like that inside the 6000 €, have a dedicated translation memory server.
Having such a tool will allow us to merge data from multiple sources.
I feel like this translation memory server will move to our Infrastructure someday and/or will be merged with other open-source projects. This is part of the l10n objectives we discussed at flock, I'll write about it to give more context, but not in this thread.

Measuring how successful we are shouldn't be too difficult (we could even ask weblate to produce these graphs, only the categorization looks missing to me). Fallback scenario will always be to self-host it.

So what would be the definition of success? Maintaining the status quo? Increasing contributors by X? Increasing translation percent by X?

I feel fine to say we'll measure the positive impact it has on the community and invite the council to decide to continue external hosting or not on a regular bases (once after a year and then every two years?). I shared some numbers so we can compare if needed. All the database dump can be shared.

But I'm not comfortable with setting numbers to define success for the Fedora Localization Community. We could easily have a situation where one of the target is not achieved, and at the same time having enabled more content to translation and new languages. This would lead to useless dispute about the value of activities compared to an indicator that looks fine. And the world of languages is complex, people will easily get hurt or argue.

I would suggest a qualitative evaluation. This really sounds like an interesting responsibility for the future FCAIC to handle this evaluation.

But I'm not comfortable with setting numbers to define success for the Fedora Localization Community. I would suggest a qualitative evaluation. This really sounds like an interesting responsibility for the future FCAIC to handle this evaluation.

A qualitative evaluation might be reasonable. I agree that trying to measure "squishy" things with hard numbers can be troublesome, but I'd like to see your proposal on what a qualitative evaluation looks like. It's not the future FCAIC asking for the money.

You do a good job of explaining the background, but what I'm not seeing is the expected outcome of this. "The translation community continues to exist" is a reasonable answer here, although it's hopefully better than that. When someone asks for hackfest funding, the expected outcome is part of the evaluation. That's all I'm asking for here.

And to be clear, I'm not saying that falling short of the stated outcome would be cause for rejecting funding in future years. Efforts fall short of their goals all the time, and that doesn't mean we give up.

My hope is that a suited tool for our needs leads us to see the l10n commun=
ity to handle meetings, taking initiatives, being able to answer translatio=
n requests from other teams and being really part of Fedora processes=2E
This would be a way to measure it=2E

That's also why I take a lot of time to have collective discussion before =
migrating and always are looking for help=2E

To be honest, changing the tool alone won't directly help=2E But having a =
collective migration project with discussion about our issues, what we want=
, what we need and having council support should help to revive us=2E That'=
s my personal belief=2E I hope it makes sense :)

@jibecfed thanks, that's a good answer :-)

Proposed voting item: 1) @jibecfed shoudl begin a cross-project discussion about Weblate; 2) 6000 EUR is allocated to be spent assuming the cross-project conversation goes well. Authorization is by FCAIC and will be sent back to council if something unexpected comes up.

Metadata Update from @bcotton:
- Issue tagged with: spending-request, ticket-vote

2 years ago

Metadata Update from @bcotton:
- Issue priority set to: Next Meeting (was: Needs Review)

2 years ago

Metadata Update from @bex:
- Issue untagged with: spending-request, ticket-vote
- Issue priority set to: Needs Review (was: Next Meeting)

2 years ago

Metadata Update from @bex:
- Issue tagged with: spending-request, ticket-vote

2 years ago

Metadata Update from @bcotton:
- Issue close_status updated to: approved
- Issue status updated to: Closed (was: Open)

2 years ago

Thank you all for your support.
I understand the FCAIC is the one to discuss with the details, to ask questions and move forward.

FYI, our project follow-up is: https://teams.fedoraproject.org/project/translation-platform-migration/timeline you're welcome to follow/contribute/share this link to anyone wishing to help.

There is a bug preventing me to contribute the existing project: https://pa=
gure=2Eio/fedora-infrastructure/issue/8116
After waiting two weeks, I decided to create a new one and to allow everyo=
ne to contribute by default=2E

Login to comment on this ticket.

Metadata