Fork me on GitHub

Has anyone else done anything structured around validating components’ configurations? My usual approach is to validate the config values types before building a system, but to defer any run-time validation (e.g. the server is up and accepts our credentials) to the start method, where I just raise for problems. Needless to say, this doesn’t present a good UX. I’m trying out a Validate protocol now, where components can report anything that stands in the way of them being able to start, with a utility fn that walks up the tree collecting errors as far as it can go.


This feels like a common problem, so I’m curious what approaches others might have used.


We’re about to go down the path of spec’ing our components so we may be wandering about in that territory shortly, but I don’t believe we have much possibility of "bad configuration" in the way we set things up so I don’t know what our experiences will be — or whether they’d be relevant to you @donaldball


I’m thinking more along the lines of “password is not accepted” than “password is not a string”, though both are germane problems to present to the operator