I use git submodules and just have nixpkgs as a submodule of my config. That way I can easily tweak nixpkgs code locally and create new PRs to push upstream.
Updating is then easy because it’s handled by git.
Actually I have 2 submodules for nixpkgs (one for stable and one for unstable).
Downside of this is that the original clone I quite slow, but subsequent updates are speedy.
I saw people pinning to a particular commit via hash too (and then you just have some helper that auto-updates that hash).
I never used channels which have this state outside of config which is not great.
Not sure why people try to discredit the question. I think it’s a great question and good to have an answer for.
People here are saying that 100ms is not noticeable latency. Believe me that it is very noticeable. When I’m streaming games via local network from PC to my tv, the. delay is ~10ms (10ms is each way so together 20ms) and It’s really easy to notice that when compared to the real thing.
And I’m not sensitive to this. For example I’m not able to tell the difference between different mice. But my friends can distinguish between them even when the difference is a few ms.
So my take would be that it depends. I would not expect Framework 16 to be the greatest gaming PC, but it should be good enough for casual gaming. I’d personally be fine with that delay.