Interesting contribution :) I think we should probably name the option --inetd or something less generic than --stdin though, as the latter is quite unclear (it begs the question of "what about stdin?").

I was also considering hooking it up as "stdin" argument to --listen. There is also the option for going for the full socket activation dance, but in my case that would actually be more work to use.

Using --listen stdin would make sense if we're sure that there's no other change in how rest-server works when put into inetd listening mode. What I mean is that if there's other factors that change besides to what/where it listens, then it's probably better to use a dedicated option like --inetd instead, as --listen stdin only tells part of the truth in that case. Let's see what others think.

We've had a discussion about this an decided not to proceed with this PR. We can easily add systemd socket activation (I've got the code already) if needed. All new features we merge need to be carefully balanced for the resources we have, as we will indefinitely maintain them.

Therefore I'm closing this PR for now. Please feel free to add further comments. Thanks!