newhinton / Round-Sync

An android cloud file manager, powered by rclone. Visit https://roundsync.com for more information!
https://roundsync.com
GNU General Public License v3.0
1.27k stars 50 forks source link

Hey you! Yes you. I want your help and ideas! #70

Open newhinton opened 1 year ago

newhinton commented 1 year ago

Hello everyone!

I need help.

And you are the one who can help me!

This project has a lot of todo's that i like to tackle in the future, and i hope that i can find someone who will help me out occasionally, to speed up development, and make maintaining this app more sustainable. In the past, two projects preceeded this app, namely rcx and rcloneExplorer. Both eventually fell silent due to lack of maintenance. To prevent this happening to this app, your help is needed! And there is something to help out for everyone:

But lastly, there is one big thing: The Name™

As a user rightfully pointed out, extRact is not really a good name that is easy to recognize or find. Maybe you have a great idea for a fitting name? Join the discussion here!

Thank you very much for your time and help, and keep syncing!

Creekie1337 commented 1 year ago

For translations, maybe Crowdin? It's probably my first choice (I think they allow free for OSS)

newhinton commented 1 year ago

Yes, transifex and crowdin have free-tiers, weblate does not.

Also here is a related thread which discusses different services, i will add more while i investigate: https://forum.antennapod.org/t/moving-the-translation-effort-to-weblate-or-another-open-source-translation-system/1181/8

alensiljak commented 1 year ago

I've had some experience with CrowdIn for an open-source app. The experience was pretty good. But I'm actually commenting since I've noticed that it is becoming more common to simply use AI to do automatic translations instead of waiting for humans to translate the next texts, then incorporate that into a future release, etc. So, just an idea worth researching. I'm pretty sure ChatGPT, for example, can translate anything you want into pretty-much any language there is.

This would shorten the release cycle and would include the translations from the moment new text is created. Contributors could still improve/change translations if something is wrong.

For example:

newhinton commented 1 year ago

Translations will be a huge task in the future of this app. Due to some issues i even had to remove some already translated elements. However, i am planning on translating even the rclone-output, but this will require a whole new abstraction-layer to make work, and therefore is only scheduled for later.

Using LLM's to translate text sounds great in theory, but in practical terms i am a bit hesitant. The reason for that is a) reliability (that is something i cannot assess) and b) legalese. The first can probably be safely "ignored" (users will probably correct the app if they find broken strings) but the second one, not so much. I don't know if i am allowed to release chatgpt's output under the gplv3, that is something i need to figure out first.

And then there is the whole unsolved issue of who did the work (as in did the people who made up the dataset chatgpt was trained on actually translate my app or are they disconnected from the output).

I will have to do some research and thinking on how i want to approach that topic.

Nonetheless, this was a good idea and i will seriously consider it! Thanks! And thanks for the other kind contribution ;) It is appreciated!

alensiljak commented 1 year ago

And then there is the whole unsolved issue of who did the work (as in did the people who made up the dataset chatgpt was trained on actually translate my app or are they disconnected from the output).

I agree and understand. This is a much bigger question and will be answered by the society. In my opinion, however, the odds are that these models will be allowed, in the end. Just like Japan did. Most, if not all, the material used for training is in public domain. If not, then they will quickly switch once the court cases start. Besides, stopping the training of the language models by imposing the copyright on the training data would greatly hinder the development of the field. And in very specific regions - i.e. you can't do that in the US but can in Japan, China, or you-name-it. This gives the strategic advantage to the countries/regions that don't block such operation. Considering that the AI is currently being implemented in the US government institutions (by Microsoft, using the above-mentioned GPT models), I doubt they would shoot themselves in the foot. That whole question looks to me like the horse owners complaining about the motor cars using the roads they built. It is doomed to failure, as this is a much bigger technological process of innovation and evolution.

On the other hand, the benefits are pretty obvious.

Just my $0.02.

newhinton commented 1 year ago

Oh sure! I don't oppose the technology, my concerns are primarily moral ones. I dont want to take advantage of others work if they did not consent to it, and for LLM's i just have not given it the appropriate consideration until now.

While i have used it in experiments, i haven't actually used ChatGPT in a "real" project, especially one that has a commercial part. (I have decided to make the google play variant a paid one, fdroid IS free and always will be)

It is just that i have to give this a good thought before doing it. (Also, there is probably something in the terms of service of chatgpt for commercialization, but there are open source llms which could be fine.)

When i am done with the current milestone (overhauling the ui and releasing the app) translations will be the next one, and then i will figure all that out ;)

Roki100 commented 1 year ago

Hey, i can translate the app to polish, although i would also like to suggest: 1) a community discord server or matrix/telegram/whatsapp/signal channel

2) update the readme as most of it links to the original project that went silent, e.g. google play or fdroid links

oh and also onedrive is broken

Cwpute commented 11 months ago

Hi, i've used both Crowdin and Weblate as translation platforms. Although being proprietary, Crowdin provides free hosting for opnsource projects, and their interface is very nice and efficient, even on mobile Weblate is very good too, but is a bit behindqin terms of UI and usability, but that's to be expected (unfortunately) for foss solutions. It's still a very good tool :)

I'd recommend using Weblate, to strenghten the bounds of the open-source community! but i'm only talking from the translator perspective, so you might consider things differently yourself.

yurtpage commented 8 months ago

Please check the Weblate integration. The strings.xml contains a lot of strings like Tasks that are absent in the Weblate https://hosted.weblate.org/translate/round-sync/round-sync/en/

I added a translation to a new lang but it looks like the Weblate can't send you a PR. The RoundSync is really needs for translations because it supposed to be used by not technical people too. Today this is more important than adding new features.