Closed xmichele closed 2 years ago
I actually much prefer Alpine Linux (which is leaner, has a good stability & security reputation, overall less bloat and it's less conservative than debian; furthermore I prefer the Alpine packaging experience as it's much easier). It always remains a bit subjective of course, but I see no real reason for switching to debian for now, the only disadvantage we currently have is that we have to grab pandoc from edge (it's still a stable pandoc release though), but in your image you have to grab dasel from upstream as it's not packaged for debian.
For alpine I referred to this as well https://dev.to/pmutua/the-best-docker-base-image-for-your-python-application-3o83 For dasel it's a official linux release. Anyway this is the debian slim, so for the bloat point of view we're ok. In every case, it's not mandatory but also consider the decision in a long term view, as the packages could change and it's more likely to have stable & updated on debian, apart from edge pandoc
Ah yes, I've seen that article before, I understand its points yes, that's also why I recommended using the py3-* Alpine Linux packages instead of pip where possible, those are prebuilt and checked so pip doesn't have to handle that. Hopefully in the future more wheels will be published that are compiled against musl too. Of course, this is only relevant for Python packages that are not pure python but have some native bindings.
For the docker build problem on the alpine (master branch) I mentioned here https://github.com/CLARIAH/tool-discovery/pull/18#discussion_r919790678 , try to do a docker build --no-cache of codemeta-harvester
For the docker build problem on the alpine (master branch) I mentioned here CLARIAH/tool-discovery#18 (comment) , try to do a docker build --no-cache of codemeta-harvester
Hmm.. let's further discuss that in that thread. I can't seem to reproduce that problem.
debian:stable-slim docker image & stable (not test) version of pandoc. More image footprint but improved stability, performance & security respect to alpine . Suggested pull acceptance coupling with discovery-tools pull request of the same branch. It already included my improvements on the master branch