Reactions to the JOSM I18n contest 2025

[stoecker]

I started a new contest in 2025, as I'm really unhappy with the completeness of our translations:

StartupPage contains following text:

JOSM supports 41 languages beside English. But only 7 of these (German, Spanish, Dutch, Belarusian, Russian, Ukrainian and Portuguese) are above 99% completeness for JOSM core. Even such important languages as French are not. I hereby challenge the JOSM users to change that situation and let 2025 be the year of the JOSM I18n contest. See Translations (and the statistics table linked there).

If you language reached 100%, there is still our help wiki needing translations: Translations/Wiki.

Note for English speakers: When you can't translate, you can help nevertheless: There are many help pages missing or needing review and improvement, see DevelopersGuide/HelpSystem/HelpTopicsList.

This ticket should be used to track progress. For the contest I'll focus on the JOSM core translations, but I hope that also means the other places when JOSM needs I18n support aren't ignored.

I'll try to report progress here each or every second week as long as there is progress.

version	ar	ast	be	bg	ca	ca@valencia	cs	cy	da	de	el	en_AU	en_GB	es	et	fa	fi	fr	gl	hu	id	is	it	ja	km
r19277	93.9	51.5	99.8	85.3	58.7	45.1	91.6	26.3	81.8	100.0	54.0	77.6	99.8	99.8	24.9	39.9	50.0	92.3	58.6	88.7	67.4	22.7	93.9	74.2	32.6
r19278	93.9	51.5	99.9	85.3	58.7	45.1	91.9	26.3	82.9	100.0	54.0	77.5	100.0	99.9	24.9	39.8	50.0	100.0	58.7	89.6	67.4	22.7	96.1	74.2	32.5
r19283	93.8	51.5	99.9	85.3	58.6	45.1	91.9	26.3	83.0	100.0	54.0	77.5	99.9	99.9	24.9	39.8	50.0	99.9	58.7	89.6	67.4	22.7	98.0	74.2	32.5
r19287	100.0	51.5	99.9	85.2	58.6	45.1	91.9	26.4	83.1	100.0	54.0	77.5	99.9	99.9	24.9	39.8	50.0	99.9	58.7	89.6	77.8	22.7	99.9	74.2	32.5
r19298	99.8	51.4	99.7	85.1	58.5	45.1	91.7	26.3	83.2	100.0	53.9	77.4	99.8	99.7	24.9	39.8	49.9	99.7	58.6	89.5	77.7	22.6	99.8	74.1	32.5
r19306	99.7	51.3	99.6	85.0	58.5	45.0	91.7	26.3	83.6	100.0	53.9	77.3	99.8	100.0	24.9	39.8	49.8	99.7	58.5	89.4	77.6	22.6	100.0	74.0	32.5

version	ko	lt	mr	nb	nl	pl	pt	pt_BR	ru	sk	sr@latin	sv	tr	uk	vi	zh_CN	zh_TW
r19277	66.3	54.3	30.5	75.6	99.8	83.3	99.4	94.8	99.8	86.4	38.1	61.9	54.7	99.7	52.5	83.8	83.8
r19278	66.3	54.3	30.5	75.6	99.9	84.2	99.6	95.1	99.9	86.4	38.1	61.9	54.7	99.7	52.5	84.0	83.8
r19283	66.5	54.3	30.5	76.0	99.9	84.4	99.6	95.3	99.9	86.4	38.1	63.4	54.7	100.0	52.5	84.1	83.8
r19287	66.5	54.3	30.4	76.0	99.9	84.4	99.6	95.3	99.9	86.4	38.1	77.3	54.7	99.9	52.5	84.1	83.9
r19298	66.4	54.2	30.4	75.9	99.9	84.2	99.4	95.1	99.7	86.3	38.0	96.6	54.6	99.7	52.4	84.1	83.8
r19306	67.0	54.2	30.4	75.8	99.9	84.2	100.0	99.0	99.9	86.2	38.0	100.0	54.6	99.9	52.4	84.0	83.7

[stoecker]

Starting State (r19277):

de	en_GB	es	nl	be	ru	uk	pt	pt_BR	ar	it	fr	cs	hu	sk	bg	zh_CN	zh_TW	pl	da	en_AU	nb	ja	id	ko
100.0	99.8	99.8	99.8	99.8	99.8	99.7	99.4	94.8	93.9	93.9	92.3	91.6	88.7	86.4	85.3	83.8	83.8	83.3	81.8	77.6	75.6	74.2	67.4	66.3

sv	ca	gl	tr	lt	el	vi	ast	fi	ca@valencia	fa	sr@latin	km	mr	cy	et	is
61.9	58.7	58.6	54.7	54.3	54.0	52.5	51.5	50.0	45.1	39.9	38.1	32.6	30.5	26.3	24.9	22.7

[stoecker]

Seems some languages accepted my contest: The new state is

de	en_GB	fr	be	es	nl	ru	uk	pt	it	pt_BR	ar	cs	hu	sk	bg	pl	zh_CN	zh_TW	da	en_AU	nb	ja	id	ko
100.0	100.0	100.0	99.9	99.9	99.9	99.9	99.7	99.6	96.1	95.1	93.9	91.9	89.6	86.4	85.3	84.2	84.0	83.8	82.9	77.5	75.6	74.2	67.4	66.3

sv	gl	ca	tr	lt	el	vi	ast	fi	ca@valencia	fa	sr@latin	km	mr	cy	et	is
61.9	58.7	58.7	54.7	54.3	54.0	52.5	51.5	50.0	45.1	39.8	38.1	32.5	30.5	26.3	24.9	22.7

Changes:

• French people took the bait and translated everything. Yeah!
• Italian did nearly half the work to reach 100%

P.S. For everybody wanting to help: Automatic translation has become much better over the years: For me ​https://www.deepl.com e.g. speeds up the work, as in contrast to the situation in the past you now can use the translation with small modifications. Years ago it was much faster to translate text from scratch compared to correcting automatic translations.

[skyper]

add table for better comparison of progress

[stoecker]

Sort by language code.

[stoecker]

Added note about Wiki and English speakers.

[stoecker]

In 19283/josm:

see #24067 - I18n update

[stoecker]

Not much progress in the last week. Italian improved, but still did not reach 100%.

de	uk	be	en_GB	es	fr	nl	ru	pt	it	pt_BR	ar	cs	hu	sk	bg	pl	zh_CN	zh_TW	da	en_AU	nb	ja	id	ko
100.0	100.0	99.9	99.9	99.9	99.9	99.9	99.9	99.6	98.0	95.3	93.8	91.9	89.6	86.4	85.3	84.4	84.1	83.8	83.0	77.5	76.0	74.2	67.4	66.5

sv	gl	ca	tr	lt	el	vi	ast	fi	ca@valencia	fa	sr@latin	km	mr	cy	et	is
63.4	58.7	58.6	54.7	54.3	54.0	52.5	51.5	50.0	45.1	39.8	38.1	32.5	30.5	26.3	24.9	22.7

[aat765]

My apologies for the delay in providing the Arabic translation updates. Previously, I was actively involved in ensuring the Arabic translation was up-to-date and comprehensive for all of JOSM's text. I'll be consistently available going forward to maintain and update the Arabic translation for the JOSM editor. Thank you for your dedication to this fantastic software!

[stoecker]

Whoa, if not one of the others tops that, tomorrow you'll win the "update cup" ;-)

[stoecker]

In 19287/josm:

I18n update, see #24067

[stoecker]

Really good progress this week: Italian and Arabic reached the target, but last weeks most effort price goes to Sweden: +14%. Indonesian with 10% is second place ;-)

I moved translations from wikipedia and pt_assistant back to our standard Launchpad system, copied and fixed the translations. Nevertheless as Transifex was incomplete for some languages there is new work to do.

[stoecker]

No progress for most languages this week, instead the complete ones are loosing percentage again. Biggest exception: Sweden is on the finish line, thanks to Joel Grafström.

[aharvey]

For en_AU and all the other en_* variants, is the intention to have translators only fill in those where the translation is different, or do you want 100% completion even if 99% are unchanged?

As a translator for en_AU, it's easier to maintain if we only store what's different, and then fallback to en. Thoughts?

[stoecker]

Replying to aharvey:

For en_AU and all the other en_* variants, is the intention to have translators only fill in those where the translation is different, or do you want 100% completion even if 99% are unchanged?

Yes. Please. Reason: There is no other way to verify whether a text is reviewed or not. JOSM texts constantly change (usually additions, but sometimes also modifications) and different people are working on this. The only way to keep track is mark unchanged texts as "done" by copying them (with one click :-).

As a translator for en_AU, it's easier to maintain if we only store what's different, and then fallback to en. Thoughts?

The resulting translation file josm/trunk/resources/data/en_AU.lang​ (see ref file josm/trunk/resources/data/en.lang​ for all texts) does not contain identical strings, but only a marker whether it is missing or identical (----- or +++++ in the ASCII view). So there is no disadvantage of storing identical translations.

Note the above links are only for the core texts. But the wiki will display all translation files this way, but if you want to see it you need to go to each plugin individually...

P.S. Please do not use the translation to fix bugs in original strings. Rather report them in a ticket to be fixed in the original...

[stoecker]

Lots of progress for Swedish and Brazilian, so I did an early update.

[stoecker]

P.S. If someone noticed help is wanted because now a picture is there as eye catcher please tell me (A picture is worth a thousand words). I want to know if advertising helped ;-) Note that there was a text message of similar content in the last 4 weeks.

P.P.S. It's not one picture by the way...

[stoecker]

Add link

[anonymous]

Replying to stoecker:

P.S. If someone noticed help is wanted because now a picture is there as eye catcher please tell me (A picture is worth a thousand words). I want to know if advertising helped ;-) Note that there was a text message of similar content in the last 4 weeks.

P.P.S. It's not one picture by the way...

I noticed it, but I also noticed it when I only had text, and I also translate from time to time, so I shouldn't “count for statistics” or draw any conclusions.

Perhaps it would be interesting to specifically challenge users of certain languages (some 3 or 5 languages) to translate, generating specific AI images for them (animals wearing traditional clothing from their country?). But it would be necessary to know the ratio of the highest number of users editing on OpenStreetMap and the lowest number of JOSM's completed translations.

By the way, even if the challenge is over, it would be good to keep showing an image that changes from time to time. An image that makes people smile and that is linked to the maps (I liked the first one with the little animals around the laptop).

[stoecker]

Replying to anonym:

Perhaps it would be interesting to specifically challenge users of certain languages (some 3 or 5 languages) to translate, generating specific AI images for them (animals wearing traditional clothing from their country?). But it would be necessary to know the ratio of the highest number of users editing on OpenStreetMap and the lowest number of JOSM's completed translations.

Also thought about that. At least 3 in the current batch are of that style. Will see next month if I add some new specific ones. Thought the more specific the images get the more manual work is involved.

By the way, even if the challenge is over, it would be good to keep showing an image that changes from time to time.

Maybe. But only from time to time, not always. But year 2025 only started, so there is still a long contest time...

An image that makes people smile and that is linked to the maps (I liked the first one with the little animals around the laptop).

Probably the second one. First one was on Sunday also with 5 animals. For the next month there will a new one every day. I hope to catch different tastes. Thought many have animals (I like them :-)

Thanks for the help and the feedback. It's not always easy to know what's a good idea and what not.

[midgard]

I like animals, but the furry images are really rather disgusting for me though...

For me having a normal image is fine; not everyone would agree: ​https://mastodon.multimob.be/@bxl_forever/113948607877107018

[stoecker]

I would really have wondered if nobody complains with harsh words after the discussions we had with the animated stars around christmas.

[Woazboat]

I'm also not a big fan of the images, not because they are somehow 'furry' images, but because I believe generative-AI is massively detrimental for artists and the concept of art in general. I'd rather see no image or even just a simple crude sketch instead of AI generated 'slop'. A simple banner image with the josm logo and some text would also do the trick just fine if the goal is to attract more attention.

The reactions to the images that I've seen so far from other OSM community members have been mostly negative and have focused on the images being AI generated. They might be attracting attention, but not in a good way.

(This should not be taken as an insult to those who put them there, I know the intentions are good)

[stoecker]

Replying to Woazboat:

I'm also not a big fan of the images, not because they are somehow 'furry' images, but because I believe generative-AI is massively detrimental for artists and the concept of art in general.

Here I personally disagree a lot. AI will change the way art is created and it will be a very hard time for artists, but it also will enable normal people to do art themself like the internet 30 years ago allowed people first time to publish and communicate in a form like never before. And I believe (though I'm not sure) that the good artists will survive the process, the bad ones not. I'm not sure how "becoming an artist" will be in the future, because it will be much harder in the "I'm not good enough yet" phase to earn money. But here I'm also optimistic, that there will be a way. I usually e.g. buy comics (or graphical novels as they are called nowadays) when they are innovative even if the art is lacking. And AI and innovative is still some way in the future - it's easy to generate something similar to something else, it's extremely hard the leave the trodden path.

Anyway I'm programmer and that profession will also be affected. And like for the artists I'm pretty sure that the result will be to separate the wheat from the chaff. The good ones will survive, that bad ones will be replaced by AI.

I'd rather see no image or even just a simple crude sketch instead of AI generated 'slop'. A simple banner image with the josm logo and some text would also do the trick just fine if the goal is to attract more attention.

While I agree that AI generated images can be bullshit, I did not simply take any crap, but designed the images properly and threw away anything which is simply incorrect. And I tried in any image to reference JOSM and the I18N task. Thought the results are based on what I like and accepted. If the images lack ideas, that's my fault - not the AI's.

I've seen many human artist images which are much worse than what I present. I.e. body proportions of the AI images are actually very good compared to many wanna-be artists which don't get even basics right even after years.

If people reject AI images only because they are AI generated, then that's their good right, but they will have a hard standing in the next years.

The reactions to the images that I've seen so far from other OSM community members have been mostly negative and have focused on the images being AI generated. They might be attracting attention, but not in a good way.

Well, it's as easy as it is: Without AI such a type of advertising would simply be impossible. Such a minor image would cost several 100€ per image. There isn't even the slightest chance that would work. BTW this also the reason why you are sure this is AI generated stuff, because I couldn't have afforded artists to do the same for me. Otherwise the images are in the same league as I could commission somewhere.

I checked the web yesterday a bit and found positive comments, so that's the normal way it is: The people disliking something are much more expressive. I didn't expect that everybody likes that.

As the I18n situation of JOSM has not improved for more than a decade I'm willing to accept a few people who reject that approach if the result will be a few people actually helping.

For me the ugly part is actually that such a big eye catcher is needed (your suggested banner would fall in the same category). But well, everything else failed.

[stoecker]

P.S. If I'm not wrong tomorrow there will be a non furry image ;-)

[spughetti]

Replying to stoecker:
If your heart isn't in art-making, you shouldn't be doing it. AI content is soulless and lacks any originality. As WoazBoat said, even a crude sketch would've been better. Because people would know a human was behind it, and that's what gives art character. The AI content you generated both looks bad and doesn't evoke the slightest bit of joy. There's plenty of people in the OSM community that would throw together a little doodle to help out, and it would be leagues ahead of what you've added now.

[Zverikk]

Dirk, you might have missed it, but in the open-source community there is a great resentment towards AI slop. The lack of intentionality behind AI-generated images seeps into your product, and devalues JOSM and the work its users doing. Mainly that's the reason Rapid hasn't gained much adoption — because of AI being core to its values. And now JOSM greets users with AI-generated stuff: it just presents an unfavourable picture of what the editor is.

As Erin writes (​https://en.osm.town/@erin/113958462421547433 ), there are enough people in and around the OSM community who could make a pretty picture without using AI.

[stoecker]

Replying to spughetti:

Replying to stoecker:
If your heart isn't in art-making, you shouldn't be doing it.

That sentence make no sense for me. I do lot of things every day which aren't my favourite.

AI content is soulless and lacks any originality. As WoazBoat said, even a crude sketch would've been better. Because people would know a human was behind it, and that's what gives art character.

That makes even less sense. Art is a product like everything else and I either like it or not. The process how it was made has no influence. Now I tell you that one of the pictures in the next month was not AI made. I assume that at the end of the month you will be able to tell me which one?

The AI content you generated both looks bad and doesn't evoke the slightest bit of joy.

That sentence is simply wrong. Searching the web you can find people liking it, as well as people I know. If you like it or not is not the only reality which exists.

There's plenty of people in the OSM community that would throw together a little doodle to help out, and it would be leagues ahead of what you've added now.

Seems you are missing the point: this is a cry for help, as the OSM community is not providing the help needed to get JOSM I18N to a decent state.

[stoecker]

Replying to Zverikk:

Dirk, you might have missed it, but in the open-source community there is a great resentment towards AI slop.

Then maybe I'm part of a different OpenSorce community, because I appreciate AI since the development of AI stepped up from research stage and gained momentum. I see the issues it has like any new technology and half a year ago I would not have made such an approach as quality wasn't good enough. Essentially I have been waiting for that stuff since I was a little boy.

I'm sad for people who aren't able to see the potential. But in my youth I had similar discussion about the internet, which was only for freaks like me and never would be of any use to decent people.

The lack of intentionality behind AI-generated images

Now what that should be? These images present my ideas. Not something an AI decided. For example todays image: I wanted an image of an anthro mouse in the jungle holding a paper with 'Help JOSM' on it and I got it as requested (well I needed to add the JOSM which AI forgot). And that is true for each and every image. They are all based on my ideas. There are other ideas I couldn't create because the AI never got it right. It's a tool like any other tool.

Actually you are insulting me, when you say there is a lack of intent.

Essentially you're saying that people who use different tools than I for their creations are somewhat better. Makes no sense.

[impiaaa]

I also agree that the AI images should be removed. In addition to all the points above they are violating multiple CC-ND and copyrighted sources.

Actually you are insulting me, when you say there is a lack of intent.

It is insulting to users and artists to display such slop in the first place.

Makes no sense.

It makes no sense for an open-source project to reject and insult its users and contributors.

[midgard]

I suppose that a flame war about AI-generated pictures can hardly be considered on-topic in the JOSM issue tracker...

[Ceirios]

Hello

Speaking as someone working on translating, I'd like to give props to everyone who is working really hard on their languages, because Launchpad (in my opinion) is... well, let's just say, not the best CAT tool I've ever used. Or, I just don't know how to use Launchpad.

I wonder if changing software might help attract translators?

[stoecker]

Replying to Ceirios:

Speaking as someone working on translating, I'd like to give props to everyone who is working really hard on their languages, because Launchpad (in my opinion) is... well, let's just say, not the best CAT tool I've ever used. Or, I just don't know how to use Launchpad.

I wonder if changing software might help attract translators?

Launchpad is suboptimal, but ok for doing a few translations. For more I suggest to download the .po file, use whatever tool for translation fits best (there are dozens of tools for po editing, e.g. KBabel) and upload the .po file when finished. Launchpad automatically handles any possible conflicts with other editors, but I'd suggest not to wait too long between download and upload, but rather work in smaller batches.

P.S. Request download only for a single language! A request to download all of them takes hours!

[thibaultmol]

Claiming that AI images are art is spitting in the face of any actual REAL artist.

You say AI lets "normal people" create art, but it doesn’t. it just mimics it. The process matters, and AI shortcuts strip away what makes art meaningful. Saying "good artists will survive" ignores the fact that many talented artists will be pushed out, not because they lack skill, but because AI floods the space with cheap, lifeless alternatives.

The whole point of OSM is that it's a HUMAN project. Using random AI slop is also spitting in the face of those people. if OSM started getting full of AI slop, we'd also make an outcry about it. These are very bad optics

EDIT: I don't want this to be flame war, I want this to be proper discussion because this is something that NEEDS TO BE DISCUSSED. And the reality is that I made a separate issue about it but that got shot down immediately and when I didn't reply to stoecker's reply within an hour, I got called out for not replying.... https://josm.openstreetmap.de/ticket/24124 like... what? This isn't a chat, this is an issue tracker with email notifications. Don't expect immediate replies. Also some people sleep...

[gaben]

Replying to Ceirios:

I wonder if changing software might help attract translators?

I would translate, for example. Fancier tools offer translation memory, auto translate and better tooling in general.

[Zverikk]

Replying to stoecker:

The lack of intentionality behind AI-generated images

Now what that should be? These images present my ideas. Not something an AI decided. For example todays image: I wanted an image of an anthro mouse in the jungle holding a paper with 'Help JOSM' on it and I got it as requested (well I needed to add the JOSM which AI forgot). And that is true for each and every image. They are all based on my ideas. There are other ideas I couldn't create because the AI never got it right. It's a tool like any other tool.

Actually you are insulting me, when you say there is a lack of intent.

Essentially you're saying that people who use different tools than I for their creations are somewhat better. Makes no sense.

Sorry, I did not mean to insult you. I understand that you wanted a picture of a mouse, you asked AI, and you got your picture. It is indeed very efficient, and to a normal person, it's not obvious what is wrong with that. That's why our shops are full of children books with AI slop on covers. They are just pictures with funny animals!

Regarding intentionality, that's a term describing a work of art (and code is art too). Having an intent is different from having a task. In art, intent goes into every little detail, not just the general direction. AI does not do that, it does not have an intent, only a task and a lot of accumulated statistics. Ted Chiang has explained that pretty well in this New Yorker article, which I think is the required reading: ​https://archive.ph/uTnxC

So when you use AI slop on the homepage, people immediately see that you used AI and fed it a couple words. My first thought was, why not just print those words? They would be certainly worth more than a thousand AI pictures. And in addition they would save wood and clean water used for producing the picture.

[stoecker]

Replying to Zverikk:

Ted Chiang has explained that pretty well in this New Yorker article, which I think is the required reading: ​https://archive.ph/uTnxC

I only parsed the text but I disagree with the basic assumptions of this, but that's off-topic. I is not the target to do art here, it is an advertising campain. And I don't believe most people will define advertising as art.

My first thought was, why not just print those words? They would be certainly worth more than a thousand AI pictures.

No. The first thing I learned in my professional career: A picture is worth a thousand words. That may be an inconvenient truth but it has been verified again and again. And these words have been there for a month. I got a lot of feedback when I asked and often that was "Haven't seen it at all". That certainly is no longer true for the images as this shit-storm proves.

[Zverikk]

Replying to stoecker:

I only parsed the text but I disagree with the basic assumptions of this, but that's off-topic. I is not the target to do art here, it is an advertising campain. And I don't believe most people will define advertising as art.

Advertising is art. Otherwise we won't have Cannes Lions and other innumerable ad awards (I was following a professional ad director for a few years). And advertising involves choices like any other work: you could have chosen to use big bold letters, or ask the community to draw a picture. Look, Matt did, and met with universal praise: ​https://en.osm.town/@watmildon/113959532100002726 . "Just an ad" is not a reason to not give it any thought.

My first thought was, why not just print those words? They would be certainly worth more than a thousand AI pictures.

No. The first thing I learned in my professional career: A picture is worth a thousand words. That may be an inconvenient truth but it has been verified again and again. And these words have been there for a month. I got a lot of feedback when I asked and often that was "Haven't seen it at all". That certainly is no longer true for the images as this shit-storm proves.

Well, there is a saying. But it should bear the question, why? To me, when I look at art, I see intentionality in every little detail. Clothes patches in Klimt's works, bridges in the background of Mona Lisa, waves and shipworkers in Aivasovsky's sea epics. That's why I love maps and can spend hours reading them: every detail is important, every detail opens something new about the world and the art of mapmaking. That is why it is worth thousands of words.

And it does not work with AI-generated art. It is worth exactly the amount you put in. When you try to zoom in, when you start contemplating its "art", you will see everything breaking apart, blurred, botched, broken. And unlike real artists, who could do the same, but having intentionality, explain the reasoning behind those features, AI... just does that. Because its pictures, just like songs and prose, are just an average of everything it has learned. No intention, no meanings, no detail, just an average.

I think JOSM is not average. It is an amazing editor with miles of depth, thousand little details that suprise and delight me every time I study it. And having a picture that is not worth the editor on the editor's front page devalues it and the work that has gone into it.

[ceiriosm]

Replying to stoecker:

Launchpad is suboptimal, but ok for doing a few translations. For more I suggest to download the .po file, use whatever tool for translation fits best (there are dozens of tools for po editing, e.g. KBabel) and upload the .po file when finished. Launchpad automatically handles any possible conflicts with other editors, but I'd suggest not to wait too long between download and upload, but rather work in smaller batches.

P.S. Request download only for a single language! A request to download all of them takes hours!

To be honest, this sounds like a lot more effort with very little gain for translation (and from what I could see, KBabel is no longer supported and now superseded). How about looking into something like Weblate, which is still FOSS? The ability to write comments for other users and co-ordinate, search properly, do it all in one interface...

As for the AI images... I have to agree with the criticism. AI is a huge threat to the translation industry, and it's really offputting for me to see it on the MOTD and to see it be defended here.

[1ec5]

The MOTD images function as decorative clip art, even as they imitate something more profound. I don’t think much effort and imagination would’ve realistically gone into them one way or another, but stylized text would’ve caught just as many eyes without roping JOSM into the broader societal discussion about the merits of generative AI. And here I thought it was a little corny for the iD v3 prototype to ​greet the user with “☀️ Good morning!” on launch. Clearly a lack of imagination on my part.

Judging from the original ticket description, this contest is as much an act of desperation as that mouse holding up a “Help” sign. It is a plea for help, though it might not be the most effective messaging. The furry animals are staring at a community of JOSM users rather than a community of translators. Ideally, software is translated by people who possess both translation skills and familiarity with the software, but otherwise, software translation skills probably matter more. There is some overlap between FOSS users and FOSS translators, especially in European languages and conlangs. However, many languages are just as underresourced in FOSS as in the broader IT industry. Tech-savvy FOSS users are already fairly likely to tolerate software in English rather than their own native language, especially if the software is filled with jargon terms whose translations would be even more obscure.

A lack of volunteers is probably only one reason for the state of JOSM translation. JOSM is only one of a few ​major OSM projects that manage translations on Launchpad. Launchpad hosts many FOSS projects, but it’s still a relatively obscure platform for software translation. Many projects use other ​translation management systems (TMS), even proprietary ones, due to their superior translation interfaces or developer APIs. At this point, I can’t pin it on a specific issue; it’s death by a thousand papercuts. Ergonomics can significantly affect participation, because software translation is usually divided up into such itty-bitty tasks. After all, how many JOSM users reach for iD or Level0 when they need to make a quick edit?

Then there’s the workload. en.lang has nearly 55,000 words across 14,378 strings. By contrast, ​the core iD project on Transifex has only 13,692 words across 1,862 strings, excluding the strings for id-tagging-schema, editor-layer-index, and osm-community-index. This is entirely reasonable, given JOSM’s sheer range of capabilities. Unfortunately, the Launchpad project lacks any distinction between more and less important strings, not even an option to focus on one dialog or another. Coming in as a new translator is rather like visiting your hometown in OpenHistoricalMap for the first time and feeling overwhelmed by the vast, undifferentiated blankness. The motivation isn’t there for a typical translator who commits to doing everything in one sitting until they’ve gotten enough dopamine from a steadily changing progress bar. Other TMSes integrate translation memory and machine translation (Google Translate, DeepL, etc.) as on-demand suggestions, which let translators focus on the more difficult strings. Launchpad has some translation memory from other FOSS projects, but in my experience they haven’t been as helpful as the ones from other TMSes.

Those who decry JOSM’s most recent foray into automation may not realize that JOSM shipped with machine-generated slop for years. Someone once laid waste to the Vietnamese localization using a machine translator, and it took some arm-twisting to get it removed: #21720. JOSM’s reputation in the Vietnamese community took a hit from that episode. LLM-powered machine translation could yield somewhat better results, but translating short strings is difficult without context, no matter who or what does the translating. OSM has its own jargon that would make the task more complicated, and JOSM also seems to suffer from some string reuse and “Lego brick” ​string concatenation. Who knows if “For” can be translated accurately? A human translator would be quite puzzled by “For the body”, especially with the explanation “group ‘For the body’ combo combo ‘Second hand’”. I can just see the MOTD image that comes out of that prompt…

This isn’t just sniping from the sidelines. I’m trying to pitch in, but it sure feels like a drop in the bucket. Translators will respond better to a plea for help that comes with a good contributor experience, including moral support. We’ve been competing against the machines for a long time now and know how to find places where we’re valued. A thoughtful approach to soliciting contributions will come in handy the next time the project desperately needs more translators, designers, or maintainers.