Selira & Turtlex
Hey Turtlex, I’ve been sketching out a deterministic scheduler for distributed systems that uses a minimal conflict graph to reduce contention. I’d love to turn it into an open‑source library—maybe in Rust for safety and speed. What do you think of tackling that as a collaborative project?
Sounds like a solid idea—deterministic scheduling can kill latency, and a minimal conflict graph is a neat way to keep the graph small. Rust’s safety will keep the data races out, but you’ll want to pick a graph crate that gives you mutable adjacency lists without a lot of boilerplate. petgraph is a good start, but maybe look at rustc_graph for low‑level control.
Define the conflict relation first: is it symmetric? Do you need weighted edges? Once you have that, a deterministic topological sort or a consistent tie‑breaker (like node ID or hash of the job payload) will give you a repeatable order.
Open‑source is a great way to get feedback—just make the core algorithm a library with clear docs, unit tests, and a CI pipeline. Maybe start with a minimal CLI demo so people can try it against a toy workload. If you run into performance bottlenecks, we can profile and see if we need lock‑free queues or async runtimes. Let me know if you want to dive into the API design or start the repository layout.
Thanks for the pointers. I’ll start by pinning down the conflict relation—symmetry is a given, but I’ll keep weights optional for now. I’m leaning toward petgraph for the adjacency lists; it’s straightforward and has good docs. Next step is drafting the API: a `Scheduler` struct with `add_job`, `add_dependency`, and `run` methods, plus a `Conflictrule` trait so users can plug in custom tie‑breakers. I’ll sketch a repo layout with `src/`, `tests/`, a minimal CLI in `cmd/`, and a CI workflow. Once the skeleton’s in place I’ll hit the first commit. Does that sound reasonable?
Sounds solid, just be careful with the `Conflictrule` trait—make it `Send + Sync` if you ever want to run jobs in parallel. Also, expose a builder pattern for the `Scheduler` so users can tweak the graph layout before they add jobs. Once you push the skeleton, I’ll fork and start adding tests for edge cases like circular dependencies and identical tie‑break values. Good luck with the first commit—feel free to ping me when you’re ready to review the API draft.
Got it, I’ll keep the trait Send + Sync and add a builder for the scheduler. I’ll push the repo skeleton to GitHub and drop you a link once it’s there. Looking forward to your edge‑case tests. See you in the repo.
Sounds good—once you drop the link I’ll dig in and start hunting for those sneaky cycles and tie‑breaker edge cases. Catch you on GitHub!
Got it, I’ll ping you once it’s pushed. Talk soon.
Sure thing—just hit me up when it’s live. I'll be on standby for the first commit. Talk soon.
Alright, I'll let you know as soon as it goes live. Catch you then.