Wasn't Python and Postgresql intentionally created for learning purposes? I'm not saying that this project is just as important, just pointing out that "learning purposes" is not a good predictor of a useless project.
For learning purposes I created quic[0]. I wanted to strip it down to the bare minimum for running something in a "container". I got as far as having it use a preexisting network namespace. The networking part felt like a major piece of work (setting up the network in bash was tedious) so I figured I'd call it a day and leave it as is. Lots of fun and learned a lot about containers/namespaces in the process.
I was going to say it'd be even more interesting if it were OCI-compliant, but I forgot Oracle already wrote an OCI-compliant container runtime in Rust: https://github.com/oracle/railcar
Though I've never really heard of folks using OCI runtimes other than `runc` and kata/clearcontainers. (edit: I guess Chromium uses `run_oci`)
One thing to note about railcar is that it completely lacks the network stack, but it will probably be my "starting point" if someday I decide to evolve `rabbitc` into something "more serious".
12 comments
[ 3.2 ms ] story [ 17.2 ms ] threadPostgresql came from adding SQL to postgres (mostly? a research project), while postgres itself was mainly a research project.
[0] - https://github.com/archevel/quic
For those lazy to click:
How a bad idea starts:- “That looks easy... I could do that.” How a good idea starts:- “That looks fun... I should do that”
Though I've never really heard of folks using OCI runtimes other than `runc` and kata/clearcontainers. (edit: I guess Chromium uses `run_oci`)