Lambda on Lambda: Serverless Haskell on AWS
10 points by abhin4v
10 points by abhin4v
Hey I was there! If Jack or Mike are here, thanks for the talks!
I had to zip off quickly at the end so I didn't ask a question that was on my mind. So on the off chance Jack sees this: what problem does the two-workspace setup solve? You recommended it a couple of times but if the specific reason was given, I must've missed it.
It avoids a circular dependency you'd hit if you tried to bootstrap a single workspace. AWS won't let you create a container-based Lambda Function without pointing it at some image somewhere, so if you wanted everything in one repo you'd have to do something like make all of your Lambda Functions and everything that depends on them conditional on a variable being set to a tag instead of unset (to see how annoying this gets, look at all the uses of count in workspaces/03-lambda-functions in the demo repo).
Even if you do that, you would still have to apply the workspace to create the ECR Repositories, then push images, then re-apply the same workspace until it has everything you want in it. That feels really weird to me, and I'd much prefer having individual workspaces without conditional resources that you can apply once (unless you're actively developing and updating things, of course).