I realize I'm just some rando on the Internet, but I'm begging you please don't introduce Yet Another CI Job Specification ™
I'm sure you have your favorites, or maybe you hate them all equally and can just have a dartboard but (leaving aside the obvious xkcd joke) unless you're going to then publish a JSON Schema and/or VSCode and/or IJ plugin to edit whatever mysterious new thing you devise, it's going to be yet another thing where learning it only helps the learner with the Radicle ecosystem, and cannot leverage the existing knowledge
It doesn't even have to be yaml or json; there are quite a few projects which take the old(?) Jenkinsfile approach of having an actual programming language, some of them are even statically typed
I also do recognize the risk to your project of trying to fold in "someone else's" specification, but surely your innovation tokens are better spent on marketing and scm innovations, and not "how hard can it be" to cook a fresh CI job spec
I likely would have already written a much shorter comment to this effect, but having spent the past day slamming my face against the tire fire of AWS CodeBuild, the pain is very fresh from having to endure them thinking they're some awesome jokers who are going to revolutionize the CI space
Appreciate both the clear feeling and nuanced take here!
It’s interesting, because it’s like the problem is partly that most of the CI offerings out there are at least a little bit gross, but also the vast number of mediocre CI offerings is a factor too.
It feels like it’d be easy to convince yourself that what you’ve built is better than everything that exists already, and hey, maybe it is! But personally I wonder if we really need is a step-change here, not an incremental improvement—something that really does make build and deploy easier, and changes how we all think about it too.
My life experience has been that answering the question is almost always a matter of "easier ... for whom ... to do what?" I think CI/CD systems often run up against the same problem that programming language adoption runs into: trying to be all things to all people for all problem domains is incredibly hard
Even what I mentioned about static typing I'm sure caused a blood-pressure spike in some readers, since some folks value the type safety and others consider it "line noise". Some people enjoy the double-quote free experience of yaml, others pound on their desk about the 7 ways to encode scalars and "but muh norway!!11"
But, taking our ragingly dumbass buildspec friend <https://docs.aws.amazon.com/codebuild/latest/userguide/build...> as a concrete example, how in the holy hell did they invent some build language without conditionals?! I'm sure the answer is "well, for our Amazon-centric use case, you're holding it wrong" but for someone coming from GitLab CI and GitHub Actions which both have "skip this job if today is Tuesday" it's a pretty glaring oversight
Hey, we are having fun doing it :) It is already pretty nice for testing things and show us the results. It is not going to be forced on users at all and it will share parts of the code with all the third-party CI integrations that we are working on.
I realize I'm just some rando on the Internet, but I'm begging you please don't introduce Yet Another CI Job Specification ™
I'm sure you have your favorites, or maybe you hate them all equally and can just have a dartboard but (leaving aside the obvious xkcd joke) unless you're going to then publish a JSON Schema and/or VSCode and/or IJ plugin to edit whatever mysterious new thing you devise, it's going to be yet another thing where learning it only helps the learner with the Radicle ecosystem, and cannot leverage the existing knowledge
It doesn't even have to be yaml or json; there are quite a few projects which take the old(?) Jenkinsfile approach of having an actual programming language, some of them are even statically typed
I also do recognize the risk to your project of trying to fold in "someone else's" specification, but surely your innovation tokens are better spent on marketing and scm innovations, and not "how hard can it be" to cook a fresh CI job spec
I likely would have already written a much shorter comment to this effect, but having spent the past day slamming my face against the tire fire of AWS CodeBuild, the pain is very fresh from having to endure them thinking they're some awesome jokers who are going to revolutionize the CI space