Use Weblate for translations

[comradekingu]

The feedback-loop from Transifex over to GitHub doesn't work, because there are no links to strings.
It is impossible to see multiple languages, suggestions are made based on deleted strings, the comment system is broken,
the messaging system is broken, the UI is broken, etc.
The terms and services are a further insult to injury, and this is preventing people from contributing outright.
It is not a good look for a libre project, and quality suffers. This is not just restricted to the translations I put care and effort into.
A move seems overdue at this point.

https://docs.weblate.org/en/latest/admin/continuous.html?highlight=webhook#github-setup
https://docs.weblate.org/en/latest/vcs.html?highlight=github#github-repositories

TL;DR Set up shop on https://hosted.weblate.org/hosting/ or self-host Weblate. Happy to help. :)

[ckerr]

TBH I don't know enough about the translation process to have an informed opinion on this. From some quick Googling it does look like Weblate has a nicer translator experience.

@mikedld wdyt?

[comradekingu]

Yes. When translators aren't subjected to the TX terms of service, and the community isn't held hostage over enterprise plan negotiations, everything seems to pull in the same direction.

https://hosted.weblate.org/projects/libretorrent/#information is doing well, and PicoTorrent was doing great on a self-hosted instance before seemingly lapsing on the domain recently.
There is also https://hosted.weblate.org/projects/nicotine-plus/ and maybe more I am forgetting.

All the available checks are beneficial, the ability to see more languages and context at once, and so on.
With Weblate it is easier to make source changes either directly, or with links to the actual files.
I did try to help qBittorrent while I used Transifex, and it isn't doing well on the linguistic front.

The current Transmission situation is "There is no activity the last weeks", so it can't hurt to abandon the lesser platform.

[mikedld]

One drawback of using Transifex some years ago (in 2017, when we switched to it) was the lack of support for .stringsdict format in their free offering, but they made it available 3 months later. Other than that, I can't think of anything else that was missing for me, but as well as @ckerr I'm not that involved with translation process to tell the difference.

The other thing is that we're lacking translation maintainers, and I'm not sure if Weblate will fix that; that might partially be the reason behind reduced activity, as I'm basically the only one who approves the join requests, and I'm not doing a great job at it. OTOH there's quite a lot of people with access there as it is, so my guess would be that no activity means that there's not that much to translate/fix for the languages where we do have translators.

As for integration with GitHub, that was a conscious decision on my part since the format and/or order used by Transifex when exporting via tx pull (and I'm guessing that would've been the format if it was to push to the repo on its own) differs from the one that's in place when importing via tx push after updating the files from code via lupdate or xgettext or genstrings, so I have a script to sync translations manually instead. I'd be happy to try out Weblate in this regard, but even then I'd be hesitant to allow a 3rd-party system write access to the repo and would rather schedule a periodic sync on our side in a more controlled environment.

See also: #100
CC: @hanklank @piotrdrag (involved in that old discussion)

[comradekingu]

@mikedld I can "maintain" it, which is to say answer question and make sure everything is good.
Approving translators isn't beneficial in my experience. The reason there are so many drive-by translators on Transifex is because it is hard to get anything finished, so people give up. The problem is sought fixed by having said roles.
It all scales better if everyone can contribute when they want and find the time, rather than 3 days later when the request gets approved (by someone unlikely to speak the language).
As for integration, https://docs.weblate.org/no/latest/admin/continuous.html#continuous-localization has a few different methods.

A self-hosted Weblate is less third-party, but that isn't where my concern lies.

[ckerr]

TBH having a volunteer who can "maintain" it would on its own make it worth moving. Every once in awhile there are PRs or questions about translations that I don't know how to answer, e.g. #2540 says

Some improvements in Turkish strings, that are not available through Transifex.

And I don't know enough about the domain to know why these strings aren't available in Transifex. Having someone to ping for i18n questions would be great.

[comradekingu]

@mikedld If you add "kingu" in https://hosted.weblate.org/access/transmissionbt/#users I can help.

[ckerr]

@mikedld I also would like permissions 😸

[mikedld]

Are we just planning on testing it out at this point or is it a general consensus that we need to move? Right now it looks to me as if nobody really cares as nobody else is joining the conversation, and I'd really like to get input from other translators... Maybe opening a thread on Transifex will help.

And I don't know enough about the domain to know why these strings aren't available in Transifex.

Switching platforms won't help with that, the issue is with strings that are in the UI (*.xib) files that aren't supported by either one. What will help is finishing the switch to base internationalization and auto-layout (#270) which will move the strings out of *.xib files and into *.strings (and hopefully *.stringsdict if we want proper plurals) files.

[bitigchi]

Translator here. +1 for the move, but more urgently, macOS side needs the layout fixes before this I believe. I guess it would be unfriendly for non-developers having to surrender a sizeable chunk of storage to Xcode just for translating Transmission.

[comradekingu]

Not so sure "GPLv2-only" is better than "later" in https://hosted.weblate.org/projects/transmissionbt/#information even though it isn't specified one way or the other in the repo.

[mikedld]

These are not the final settings, and I'm not actively migrating anything at this time. Had to create at least one component, and specify the license, to apply for the Libre plan. But note that the approach right now is to be explicit about particular license versions we use, as seen in various source file headers, and there's no option to select multiple licenses in Weblate; for the justification on this, I believe @ckerr has the context (I remember it quite vaguely so don't want to put words into anyone's mouth).

[comradekingu]

@mikedld Work done is still licensed by its creators, not by project admins.
If it isn't specifically "only" anywhere else, I don't see a reason to use it for the translations.

It may be used under GPLv2 (SPDX: GPL-2.0-only), GPLv3 (SPDX: GPL-3.0-only),
or any future license endorsed by Mnemosyne LLC.

↑ Put this in the info box and all is well.

Going GPLv2-only ruins the entire concept, and locks translations to a license without a cure clause, etc.
The imported work from Transifex already has either this project license, or something specified there.

The effort itself has already migrated. Making use of it is another matter.
I fixed all the formatting errors I could and redid the nb_NO translation.
French and Turkish has seen work by top flight translators.
Meanwhile at Transifex → "There is no activity the last weeks".

[mikedld]

Again,

These are not the final settings, and I'm not actively migrating anything at this time.

Whatever changes you (or anyone else) are doing will not be preserved, because translation at Transifex is in fact happening. The fact that you don't see it for some reason is another matter.

[ckerr]

But note that the approach right now is to be explicit about particular license versions we use, as seen in various source file headers ... for the justification on this, I believe @ckerr has the context (I remember it quite vaguely so don't want to put words into anyone's mouth).

Sure. The current text of

It may be used under GPLv2 (SPDX: GPL-2.0-only), GPLv3 (SPDX: GPL-3.0-only), or any future license endorsed by Mnemosyne LLC.

is to avoid pre-approving future licenses, but still make it easy to include them after they've been vetted. If future versions are in the spirit of earlier GPL versions I expect Mnemosyne would endorse them.

[ckerr]

Translator here. +1 for the move, but more urgently, macOS side needs the layout fixes before this I believe. I guess it would be unfriendly for non-developers having to surrender a sizeable chunk of storage to Xcode just for translating Transmission.

That auto-layout PR #270 appears to be dead; the submitter hasn't updated it for years. There is not currently an active developer on the Mac client, so just to put this out there -- if anyone wants to volunteer and pick up this work, you are needed 😄

[DavidMunch]

Another translator here: Transifex's UX is not very good. I whole heartedly approve of a move, even without not knowing Weblate. It can only be better.

[pataquets]

My 2c: user of both services with preference for Weblate. Using only web interface with no other tooling. Thumbs up here, ready to help.

[atriwidada]

Translator of several projects on Transifex & Weblate. +1 for move to Weblate.

[crayxt]

Please transfer current language coordinators, if possible. New team members should be approved by coordinators before they can contribute.

[pataquets]

In my experience, approvals have caused me long delays in contributing if admins are busy or even no longer interested. I understand it might be necessary in some languages, but I'm 50/50 on it, especially languages with less contributors.

[mikedld]

From Transifex:

Hi, Mike,

So you are moving Transmission to Weblate. Were you attracted by the fact that Weblate is open source?

Unfortunately, for translators, reviewers and language coordinators, Weblate lags in so many ways: the UI is hard to work with and slows down the translation and review process, there is no translation memory suggestions, no autofill (if I’m not mistaken), it is just a less refined l10n platform.

I’d be happy to coordinate the French (fr) translation in Weblate. Please give me the required permission to review strings in Weblate.

Will you be able to keep the current string review status from Transifex?

Kind regards,

AO
French language coordinator | Localization Lab
localizationlab.org | transifex.com/organization/otf
5C8F 0C0C 4812 7771 365C 262D 47DC 6187 47A9 22D2

[comradekingu]

@crayxt Language coordinators can be made reviewers on Hosted.
The reason quality suffers on TX stems from there being no functionality to ensure it, nor community.
There is a further divide between upstream, no linking to source, or transparency allowing people to even see translations without being logged in.
In turn TX has a stopgap measure that escalates missing functionality at a base level to the idea of roles.
It does prevent the inherent TX problem of drive-by translations, but only cements decisions that are bound to be as bad.
Instead of trying to coordinate a translation betwixt people working on it, now the authority is divided into finding people to entrust with the ability to keep people from doing so. I have seen that work never.
The changes made for nb_NO on Weblate serves as proof of how well it works for Transmission.
It is worse for small languages having one person held up asking for permission, and it is also more likely that person will go missing, making the problem worse. That does not scale well.

@mikedld
Just to be clear for those playing at home, that is the user AO, not an admin from TX or similar.
Having discussed it in person, it seems to be a case of lacking experience in using Weblate.
To be clear, translations, (zen view) reviews (matrix view), coordination (explained above), leading projects and administrating and integrating (explained below), is heaps and bounds better in Weblate than TX. The inability to do things quickly, span large amounts of strings, or actually see multiple languages at once is the first showstopper.
Then the TX comment system is hidden, and the non-threaded messaging system is a pain.
If the objective is to have people not respond or reply, or find your filed issues, TX is the way to go.
In terms of integration, the tx pull/tx push is flakey, and will gladly remove all attribution from all strings.
Weblate in contrast actually credits translators down to the VCS system.
On TX it is easy to think strings are pulled and translations don't end up used, or having to recreate branches with everything attributed to the person doing it. This includes renaming locales, because that isn't supported, and TX lets people use broken or erroneous locales to make matters worse.
If it wasn't a pain to ensure quality passed for mustard before, now there is nothing to go by.
And don't actually click "review", because that locks strings.

Weblate is actually many l10n platforms, and TX is more of a datamining operation that makes uses of every single thing it can monetize. Those are not equal concerns, and they set the two apart fundamentally.
Moreover TX isn't just closed source, but an only-as-a-service offering subject to egregious terms.
In turn everyone must be 18 or above just to accept said terms.
TX also holds entire translation efforts hostage to negotiations over enterprise plan purchases.
The only reason the Localization lab uses it is because they have six-figure yearly budgets to fork over because of their funding model. If the LL leadership had the technical prowess to do so and/or listened to their members (who prefer Weblate), it too would also have moved every project over. As a control group, the LL projects that use Weblate are doing vastly better.
Far more fixes sent upstream, more and better translations.
Also, it is easier to fix failing checks, because Weblate makes that semi-easy.
I did it for all the crucial strings failing in Transmission on Weblate.

In stark contrast to Weblate being libre software and actually helping similar projects, the closed TX only allows non-commercial libre software to be hosted without direct compensation. Transmission doesn't actually qualify there, as it takes donations.
TX closed their codebase, said they believe in open source, and then changed that around.

The fact of the matter is Weblate has translation memory suggestions, machine-translation suggestions, and automatic population of either, and optional sharing of those project or instance-wide.
On TX there are memory suggestions for deleted strings, and then one has to open those and go check to make sure they aren't live.
You want consistency, that is what you have to do, for every string.
Weblate actually handles fuzzy strings correctly instead of crudely removing strings. Etc, etc.
Strings can also be copied wholesale.
As for the UI, TX has less horizontal space available for the entry field, which wraps strings much sooner.
Trying to keep track of now multiple lines with more arbitrary entry points between source and translation, is no fun.
If the string is sufficiently long, now it even scrolls too. The solution is to copy one or both over to a sane editor, and then back again.
Imagine the process of reviewing many long strings with that in mind.
Lets say you find an error and want to commit it in VCS. There is no link, and there is no direct editing.
The fix nukes all translations and users have to dig them out of a nested-menu memory, That isn't going to happen.
What happens is users either leave the longest and most complex strings untranslated, or sometimes auto-translate them.

Weblate leaves some things to be desired, but it isn't an out and out joke.

[mikedld]

https://github.com/BuddiesOfBudgie/budgie-desktop/releases/tag/v10.6

Other Changes

• Moved from Weblate to Transifex for translations

If someone has contacts there, would be interesting to know their reasons.

[comradekingu]

@mikedld Maybe @JoshStrobl knows.

[lpapadakos]

I can still visit the Transifex site. HaAs it been terminated/operations moved house or not?

[ckerr]

@mikedld is this issue still something up for consideration?

I don't think it makes sense to leave this open forever; we should either consider moving or close this issue as declined.

[comradekingu]

@ckerr As good an idea now as it was a year ago.
10+ people agree here, among those some of the very best libre software translators around.
For the year in passing, there was a ~5-fold increase in Weblate activity on the main instance https://hosted.weblate.org/stats/
There are now "Teams", to grant permissions per user, per language, per project in a sensible way.

Transifex has had the same main issues for 15+ years. Show me just one project that manages to avoid them.
I never managed to in any capacity.
Integration, management, team-building, communication, upstream contributions, translation, review, etc.
For every way of interacting with it, it fails with no upside. That being said, I haven't tried the very latest developer tools, introduced as a breaking and forced change, so no redeeming quality in and of itself.

Budgie moved to Transifex when it got forked out of Solus.
It seems to have been the decision of the main developer, and I have no idea as to why it was done.
https://forum.endeavouros.com/t/budgie-main-dev-leaves-solus/22104/22?page=2
There were problems importing it, the old Weblate instance wasn't advertised on https://weblate.org/discover/ and the new effort uses erroneous locales.
One year on, and the quality isn't there, with sporadic coverage. Exactly what one would expect.

https://l10n.elementary.io/languages/ (Pantheon DE) did the opposite, and quality improved greatly.
40 or so languages with full coverage. (Not counting changelog entries.)

[mikedld]

@ckerr, this dropped off my radar. I'll need some more time to consider it, let's leave open for post-4.0.0.

[maboroshin]

Transifex and Crowdin have access control. Transifex and Crowdin allow users to participate in translations. The advantage of Weblate is that it doesn't require users to be approved. On the other hand, this is not keep a constant quality. Users come to Weblate who use machine translation and translate "short" (video) as "shorts"(clothes). When this happens, Weblate has no access control. (Strictly speaking, the free Libre plan doesn't have this feature)

(PS: Weblate also had a user block feature in Libre plan after mid-2021.)

Transifex and Crowdin, as well as GitHub, have terms of service that prevent harassment. Weblate doesn't have this.

---

There may be one size fits all when it comes to usability. The progress rate from Weblate's project list is very easy to read. Other translation support sites don't have this.

Weblate's UI will erase the search string if there are no search hits. This is a UI that you don't see very often in many apps. I reported this as an issue in Feb. The month of August arrived with the label "Good first issue". Sometimes you have to wait for a beginner to luck out and fix it.

Other translation assistance services integrate a list of strings with the details of a single string. Crowdin and Wordpress.org have automatic translation on the same page. To use automatic translation in Weblate, you must open one page per string. Weblate's list view (Zen mode) can’t display comments or automatic translations.

[comradekingu]

@maboroshin
TX and C prevents anyone not logged in from even browsing strings.
No service actually require users to be approved.
TX and C use access control as a stopgap way to hide its underlying flaws in trying to maintain quality.
The loss is activity, and the trend for one service over another is clear. Libre software projects start on, and move to WL.

Comitting bogus translations at that level isn't a fault of the translation engine.
When malice or sub-part translations happen, it is mostly down to UI for how to fix it, for other people.
In no way does TX or C even compare there. It is a longer discussion to show how.
The rest is flagged errors, and that is a lot better than being tied up in waiting for votes to have any consistency.

The libre plan isn't the same as WL the product.
There is a difference between only-as-a-service software and software you can own.

The amount and content in the terms and services of each respective service heavily favours WL.

3.5 is the point that "prevents malice" https://weblate.org/terms/
Arguably people with no ill-intent could be prevented from malice somehow, but it seems
to be a feign concept.
WL also has https://github.com/WeblateOrg/weblate/blob/main/CONTACT.md
which in my view is how far preventing malice in a text-file goes.

WL shouldn't delete search strings that return no hits, but it also should be welcoming to newcomers in offering easily fixed problems to solve.

It is a somewhat good idea to not make auto-translation easier, and you are absolutely correct about comments in Zen.
Auto-translations are possible in bulk with plugins. They are marked as such, but not as easy to distinguish as they should be.

[maboroshin]

For example, One Weblate user translates dozens of languages. Is he a genius who also speaks 30 languages?
https://hosted.weblate.org/user/makarrazin14/contributions/?page=3&limit=100

In fact, he translated "Resume" as a history of education, previous jobs, This is a music player lol. This is a machine translation. He also translates short videos as short pants. Weblate sells the ability to control access to these users.
https://hosted.weblate.org/translate/libretube/libretube/ja/?checksum=5cbd08025a9df8db#history

This is the biggest difference from other services. In other words, Weblate is allowing this type of user. But you don't have free charge access control. I would speak equally without embellishment, Weblate appears to be banning spam with its spam reporting and detection. Perhaps as "malice".

(PS)

• SPAM by suggestion WeblateOrg/weblate#2284
• Spammer management WeblateOrg/weblate#3548
• Add way to delete spam/unused users WeblateOrg/weblate#8470

Weblate also had a user block feature in Libre plan after mid-2021.
(End of PS)

[comradekingu]

I think many of those stem from changing source strings (as is the case for nb_NO)
https://hosted.weblate.org/translate/libretube/libretube/nb_NO/?checksum=d08f3ad2dc70d609
https://hosted.weblate.org/translate/libretube/libretube/nb_NO/?checksum=44ae79803f109969
https://hosted.weblate.org/translate/libretube/libretube/nb_NO/?checksum=39f69c6429aa2dda

In translating 250k+ strings on Hosted Weblate, I have only come across this twice.
It would be a lot easier to direct or even ban the two users in question than to fix the inherent problems that
TX and C try to hide by various other means, like hierarchy and (hierarchy) voting.
Even if, what keeps anyone fror applying for a language neither the person or the language admin speaks. Nothing.
I have administered a lot of projects on all platforms, and it just doesn't work.
Other than outright banning people, which I have done just the once, I can administrate changes just fine through having
the same access as everyone else.
Ask yourself if the project admin is magically going to find the best translators in the world to admin the respective languages? Would you at least not want to start that idea on the active and functional platform?

Never have I seen a project arrive at quality translations through TX or C, and this is why nobody favours those.
There is no sense of continuous effort, and instead there is a lacklustre and broken tool in the way of even making a single pass of the stringbase. Fixing upstream strings is also a massive pain.

There is a case for being able to change both all and source strings, and that pans out through everyone being able to correct mistakes, which is what most people do. Those are most often mistakes through no ill will, and the amount of people capable of correcting them are proving their worth through selection in accruing a history of good changes and trust. You will find no such community elsewhere.

Weblate can of course also have different roles and permissions, and so can the instance Hosted Weblate.

Weblate isn't a service. Hosted Weblate is a service.
Weblate is copylefted libre software you can do with exactly what you please.

TX and C are closed source only-as-a-service, and they use that power to spy on and commercialise the user.
They also have full power to change it as they please, without anyone being able to do more than complain.
That is a big and inherent difference.

[maboroshin]

It's a fantasy, that's all. But give real examples.

It is not a string change. The accuracy of machine translation would be around 80%. This is very different from a one in 100,000 failure.

Projects that maintain high quality translations will have translation guidelines and an approval system for participation from other native speakers. or skilled translators are the main contributors. And there are no trolls. In other words, it is the opposite of Weblate. Weblate, with its complete beginners and machine translations, will frequently see mistranslations. Furthermore, there is no ban.

No, Weblate is also commercialized. Weblate's access control is only paid for. Weblate also has a paid "hierarchy"! If troll posting machine translations again, there is no free access control. For free access control, you simply need to leave Weblate.

(PS: Weblate also had a user block feature in Libre plan after mid-2021.)

[comradekingu]

I can tell you who the other person is if that makes you feel any better.
Someone thinking they know every language in the world is a far
cry from a world where inaccurate machine-translations are somehow avoided.
It isn't as if machine translations are any more accurate elsewhere.
Having more options for them is a benefit though.
You still aren't pointing to anything unique about any platform, so it is a non-argument.

The best libre software translators in the world use Weblate. I can tell you who they are too.

Bans do happen on Hosted Weblate, but they are rare.
Instead guide people into doing better work, and first hope changing their mistakes makes it back to them via e-mails they can read. Or first and foremost get notified of inconsistencies and triggered errors.
Feedback by e-mail turned on by default, as opposed to TX or C, where the default is for every mistake someone makes to keep propagating. Translatewiki does a good job of this too.

What exactly do you think prevents anyone from posting erroneous machine translations in a system of access control?
What it really does is prevent people from fixing errors and doing work when they are free to do so.

As an example it isn't convenient to sit around and wait for the language admin to approve access two days later when you found the project you wanted to contribute to for your free weekend.

As another example, two times people have gone overboard with changing things in languages they don't know, but I endlessly fix semantic and syntactic errors in other languages.
Markup formatting and proper nouns don't change between languages.

Learn the difference between Weblate and Hosted Weblate and you will discover where the great difference to the alternatives lies. One is an offering as a service, and the other is the software.
In my world "free" means something beyond gratis. However, TX and C are neither in any department.
Only offering a service at no monetary cost to non-commercial software isn't the same as a real offering for libre software.
Weblate being commercial software doesn't mean it sells its users.
I advice you to read the terms and conditions of the services you mention.

I have written translation guidelines. If your hope is that unskilled translators both read and follow them on platforms like TX or C you are in for a surprise. Weblate at least has a field for it, but don't rely on that either.
The base problem of open access or restricting it to the abilities of a language admin is the same reliance on the platform.

Hypothetically however you are correct about translation guidelines and access control fixing things, but if you actually try to put that into the real world it isn't so easy.
Tell me what project you think of as maintaining high quality translations and stringbases.

These are some similar examples to Transmission
https://explore.transifex.com/sledgehammer999/qbittorrent/ almost unusable because of how quirky it all is.
https://translations.launchpad.net/deluge/ "Structured" permissions, yet fails at even having coverage.
https://hosted.weblate.org/projects/onionshare/ This is how to do things right.
https://hosted.weblate.org/projects/libretorrent/ Good, but there are things to improve on.

[maboroshin]

In translating 250k+ strings on Hosted Weblate, I have only come across this twice.
The best libre software translators in the world use Weblate.

You bring the fantasy. Vague, ambiguous. Translators can participate on any platform. Do they participate only in Weblate? And you are basing this on what data?

You still aren't pointing to anything unique about any platform

You just don't recognize it.

	Transifex	Crowdin	Weblate
Place	USA	Estonia	Czech Republic
Projects	40,000 Nominal (2018)	166K Nominal	1,164 Nominal
Open project	16,000 Nominal	16,869 Actual	987 Actual
Users	0.5M Nominal	2.1M Nominal	0.067M 67,413 Nominal
Team	10 peples	8 Peoples	3 Peoples
Eemployees	84	117	4 +Contributors
Display of translation rate	Per peroject	Per peroject	1 page
Integrity Details	1 page	1 page	Per string
Access control	Free	Free	Free
Participation	approval	Per project	anyone
Anti-Harassment Policy	O	O	X
Leakage of personal information	?	?	Weblate#8451

• Crowdin's Advanced Search could be referenced up to page 563. It's 30 projects per page. 30*562+9=16869
• Weblate is the "Hosted Weblate". Not all self hosted server instances. Weblate's top page claims 2500 libre software projects. Weblate displays 1165 projects in HostedWeblate on the explore page. And HostedWeblate shows 1164. However, 987 projects are actually available for display. There are over 350 projects on servers other than HostedWeblate on the explore page.
• Weblate can display all translation rates at 1 page. Excellent. However, the list of translated strings is very poor. Weblate and Crowdin can display all strings and translation history, comments, etc. in 1 page via Ajax.
• Codes of conduct, including anti-harassment, have been adopted in many large projects.
• Weblate has already output personal information to Github and it is difficult to erase all of it.

WeblateOrg/weblate#779 : This is a bad example of Weblate. Weblate can't exclude machine translation which joining without approval. The Weblate administrator used Github's free ban feature to stop the appeal. And it seems that the machine translators remained. Anyone can join, no access control, no harassment prevention, Most support correspondence is returned by nijel. - This is what makes Weblate unique. Weblate is easier for complete beginners and machine translators to join.

Translations of Wordpress and its plugins maintain uniformity and quality. There are guidelines and approval to participate. Linux Mint is also so.

[comradekingu]

So you understood Weblate can show all strings in a project at once, in multiple languages and that it is impossible in TX or C.
We are done there. Nothing more to report.
Garbage in, garbage out. Access control won't help you.
Every project on Hosted Weblate can run a custom banlist, and it is very rarely used.
The quality is better because doing anything about it is much easier, with little to no access control.
Many projects on TX or C don't use access control, but the results are never good.
There is no saving grace there.

You bring the fantasy. Vague, ambiguous. Translators can participate on any platform. Do they participate only in Weblate? And you are basing this on what data?

I am one of the top contributors on each platform except for C, because there was never anything interesting on there.
Before projects moved away from TX it used to be a relevant choice compared to Pootle.
The problem you raised is exceedingly rare, and easily fixed with the tools available on any platform.

So Linux Mint, of all things, are doing something right, in using Launchpad…?
WordPress is somehow doing well, with a custom setup?
Launchpad is unusable, and it seems the WP implementation is too.
Give me one libre software project that uses TX or C with any success.
I can give you plenty counter-examples, but it seems we are done there.

You haven't actually demonstrated any advantage to restricting access.
The things you complain about are not unique to any platform, but the issues I raised are unique to
doing access control the way TX and C do it.

It seems you haven't ever administrated any projects, and I guess that deluded you.
The term "drive-by translator" was coined on TX, and that was with paid staff to do access control.
If you think Transmission can afford that or have the resources to do it you are mistaken.
You need to have a functional platform, not hide behind stopgap measures.

If you read my answer above, you would already know Weblate has an anti-harassment policy,
but you neither read or learn, but are somehow convinced a text-file will prevent harassment…
The same goes for codes of conducts. Weblate and almost all projects, including Transmission
do without one and it is just fine.
Meanwhile they create lots of problems, but I guess the hypothetical nature of your beliefs guide you.

You point to a solved issue, that points to another solved issue.
And it says
"On the other side, none of hosted projects complained on the openness so far, so I believe it works fine for them."
but you have a problem with what exactly?

Understand that Hosted Weblate isn't the same as Weblate the software.
What do you think happens when you send support questions to TX or C?
I know what happens when you send 5-figure dollar amounts to TX yearly, and the support sucks.

[maboroshin]

I am surprised at your assertion that there is no demand for the features implemented. I see that there are about 15 times as many projects in Transifex and Crowdin as in Hosted Weblate. Chances are it will contain more successful software than Weblate. Linux Desktop Xfce is in Transifex, It will maintain a 95% translation rate in approximately 40 languages. In Crowdin, TeraCopy has 28 languages. Crowdin's Minecraft has 37,452 people participating in translation. This is more than half of all Hoted Weblate users.

• Explore Transifex
• Explore Crowdin
• Sorry, Weblate does't automatically generate a page listing successful projects.

Weblate doesn't have an anti-harassment policy. Spam and completely irrelevant text submissions seem to be dealt with. #2538 (comment) The Code of Conduct is on GitHub. Don't need it for individual projects. It's on most of the big web services. See A Code of Conduct for Open Source Communities Violence such as the F-word could be banned by reporting it to support. I don't see any issue on Weblate that addresses this.

[comradekingu]

The features implemented doesn't amount to a bunch of access control.
That access control is next to useless in garnering quality crowd-sourced translations is not a well-kept secret.
You still haven't learnt the difference between libre software and proprietary software.
What in the world is "TeraCopy"? Find something someone has heard of.
Minecraft doesn't need 37k translators, that is just a detriment.
Looking through even the "Featured projects" and "Showcases" there is nothing there.
It is abandonware, proprietary software, stuff without activity, or projects that have moved elsewhere.
GitLab remains on C, when GitHub has abandoned ship already.

That the software is successful doesn't mean the localization will be. Case in point, the GitLab translations suck.
What you get is a representation of how good of a tool the translation platform happens to be.
You didn't come up with a single project using C or TX successfully, and I haven't seen one yet.
There is an endless stream of projects with opposite results.

Weblate has an anti-harassment policy. Moreover, harassment is illegal. You evidently don't need a CoC for anything when the most diverse community on earth does a whole lot better than the other offerings.
Words are not violence. Swearing hasn't been a problem, nor do I see why it would be.

So Xfce uses TX and has bad results (I have worked on it, so I know).
KDE, LXQt, Ubuntu Touch, Trinity, Pantheon and so on uses Weblate with good and great results.

[Pentaphon]

Any progress on this? Weblate seems to be more FOSS-friendly.