Bootstrap diagram

[modship]

Here is the current connection diagram for the bootstrap refactor of #3554 :

@Ben-PH

Update with the fix of updater loop (deprecated) :

deprecated :

[AurelienFT]

I think for me the main loop/listener should be either regroup (as there is only a listener management in the main loop) or the listener should create the thread himself. I find this message passing a bit too big overhead. The verifications could be done in the thread of the client or just before as it's read only with a RwLock it will not prevent to have differents threads checking the same data.

[Ben-PH]

I love the idea of that. I'm just curious about how it's done in a way that doesn't spam syscalls, and doesn't block the polling of the stopper.

Assuming that's done, it's pretty ideal. I wouldn't be surprised if that assumption is upheld, but I don't want to fly in blindly.

The other approach I was going to try, is to use std::sync::Condvar, and use that to manage thread-progress. Mio will be much nicer though.

[Ben-PH]

...I've had a look into it. I took the tcp server examples and added some printlns... no spamming. Will use mio

[modship]

...I've had a look into it. I took the tcp server examples and added some printlns... no spamming. Will use mio

Nice 👌🏻

[AurelienFT]

It's used by tokio so I think it's optimize but didn't go in depth

[Ben-PH]

Yeah, I cloned the repo, added some printlns before the syscall!s, and saw that we're good.

I'm surprised that crossbeam doesn't bring it in to play, but that's another discussion.