I’ve long held the position that our tools are so often ahumanist junk because we’re so deeply beholden to a history we don’t understand, and in my limited experience with the various DevOps toolchains, they definitely feel like Stockholm Spectrum products of that particular zeitgeist. It’s a longstanding gripe I’ve got with that entire class of tools, Docker, Vagrant and the like; how narrow their notions of a “working development environment” are. Source, dependencies, deploy scripts and some operational context, great, but… not much else?
And on one hand: that’s definitely not nothing. But on the other … that’s all, really? It works, for sure, but it still seems like a failure of imagination that solving the Works On My Machine problem involves turning it inside out so that “deploy from my machine” means “my machine is now thoroughly constrained”. Seems like a long road around to where we started out but it was a discipline then, not a toolchain. And while I fully support turning human processes into shell scripts wherever possible (and checklists whenever not), having no slot in the process for compartmentalized idiosyncracy seems like an empty-net miss on the social ergonomics front; improvements in tooling, practice or personal learning all stay personal, their costs and benefits locked on local machines, leaving the burden of sharing the most human-proximate part of the developer experience on the already-burdened human, a forest you can never see past the trees.
This gist is a baby step in a different direction, one of those little tweaks I wish I’d put together 20 years ago; per-project shell history for everything under ~/src/ as a posix-shell default. It’s still limited to personal utility, but at least it gives me a way to check back into projects I haven’t touched in a while and remind myself what I was doing. A way, he said cleverly, of not losing track of my history.
The next obvious step for an idea like this from a tool and ergonomics perspective is to make containerized shell history an (opt-in, obvs) part of a project’s telemetry; I am willing to bet that with a decent corpus, even basic tools like grep and sort -n could draw you a straight line from “what people are trying to do in my project” to “where is my documentation incorrect, inadequate or nonexistent”, not to mention “what are my blind spots” and “how do I decide what to built or automate next”.
But setting that aside, or at least kicking that can down the road to this mythical day where I have a lot of spare time to think about it, this is unreasonably useful for me as it is and maybe you’ll find it useful as well.