Dear all, Please, don't get me wrong. I'm a long-time Debian user and Open Source project maintainer. In my own projects I continuously struggle with the fact that an improvement in one project needs to be taken over to all the other projects.
If you manage several (free software) projects yourself you may know this feeling: You try to improve the state of a repository, and you know you've done it before hence you want to avoid reinventing the wheel. But now you're unsure where the last place was where you had your latest and greatest solution (call it, "current best practice"). With that said, I know the tutorials on Debian packaging but I'm interested in actual, living implementation in popular projects, because that's where you learn current practices best from. People may perceive this as rude, but it certainly depends on how harsh or welcoming you want to be. I learned how to fish years ago, now I want advice from fishers. > What I would recommend in 2024 is that you package your software as a > private .deb — which is fairly straightforward, even if you have > binary blobs that aren't built during the package build — and then > install that inside a suitable base debian container image. I appreciate your advice, Michael. The point is, I didn't plan to come here to bore people with discussing my infrastructure implementation strategy. I don't think it belongs here. Using system packaging is exactly the intermediate step I want to use for going towards a cloud-native deployment for a nowadays traditionally deployed, VM-based setup. Intermediate steps are needed in environments that are reluctant to change. And not getting root-access is one component of the first equation of that change; having to cope with pre-installed components on the managed VM is another. But we're leaving the ground of Debian (and Open Source) here by far. I had hoped someone would point me to a good example of a popular project. One that delivers the final JAR, and a Debian maintainer takes care of packaging only. Maybe because the project doesn't care about having a Debian package at all. That would cut out the build part (Maven). I started my research with Geogebra, didn't find traces in their GitHub repositories (and it looks like they stopped promoting packages for Debian at all) and didn't know where to look for the original packaging code on Salsa. (I found it in the meantime at https://salsa.debian.org/java-team/geogebra) The colorpicker project (https://salsa.debian.org/java-team/colorpicker) seems dead-simple, but it builds from source. Is that good enough? The last code change is from 12 years ago. Thank you for your understanding, Peter
