Updates to the benchmark

[K0-RR]

Hi, I have no idea how much time-consuming this would be, but would You consider benchmarking dosr with other sudo alternatives, like doas, run0 and pleaser?
I understand this might not be that important, given the true problem RootAsRole is fixing, but it's still something that's a nice additional comparison.

[LeChatP]

Thanks for interest on the tool!

run0 acts primarily as a client for systemd dbus calls, so the performance bottleneck likely stems from Dbus overhead rather than the tool itself.

Having reviewed the please codebase, I noticed significant allocation overhead; in contrast, RaR does handle allocations way more carefully.

But indeed, testing performance with these tools might be interesting.

Currently, I'm focused on implementing new features and resolving dependency issues. The current roadmap is:

1. Implementing pipe communication with processes for more control, which is quite a big thing, so I'll set it as a 4.0 objective.
2. Implementing sgbd, as a 5.0 update, As for today, I didn't found the candidate ORM for RaR today.

The dosr binary by itself has remained quite stable with few bugs reported so far. Indeed, the xtask code is bugged, but it was rushed; I admit it was just a helper tool for my students who cannot align more than one command line...

Additionally, because Rust has evolved significantly since 2021, I'm refactoring parts of the project to leverage modern language features for better performance and maintainability. I recently updated the project to the 2024 edition and I'm using Clippy in pedantic mode for better maintainability (and verifying details that I missed).

[hiSandog]

A fair comparison should separate cold process startup from warm authorization, run the same no-op command/policy, and report distribution/variance rather than one timing. run0's D-Bus round trip is part of its architecture, so include it while also labeling where time is spent; caching and authentication prompts must be controlled consistently. Pinning tool versions, kernel/systemd, hardware, and benchmark script in the repo would make future updates reproducible.