This is specially important, where in the new releases on NixOS, it usually takes time for all binaries to be available. In the recent upgrade, my less powerful machine, couldn’t catch up with all the builds, and the failures happening every few hours proved to be frustrating.
On the framework machine, it took a few hours to complete though. So I’m thinking, of having a binary cache on any of my machines, and exposing them on the network to use as substitutions of the builds, will help with:
- Reducing the download time.
- Reducing the build time.
Currently I’m using a modified version of Sunaku’s awesome keyboard layout on my Glove80. The main improvement for me, apart from having the ergonomy, is the home row mods. Yesterday, while reading Uses This, I learned about Keyd, and man, how did I miss this?
It seems like, it enables the same functionality of framewares like ZMK, on the normal Linux setup, and it’s agnostic from the GUI. This sounds promissing.
This is specially important/interesting for me, given the fact I literally dropped my Evil like bindings from the Emacs setup, having the comfort of the Glove80 setup. However, for <10% of the times, I’m on my laptop, on the move, and having this, I don’t need to crack my muscle memory to do Ctrl/Alt/Shift with the normal positions.
- Enabling the setup needs to be global and system level. Consequently it not only covers the GUI desktop, but also works out of the box on TTY.
- The response time seems more effective than what I feel on Glove80.
- While setting up, it’s quite important to not make a mistake on
replace keys for
overloadt2
’s second argument. Otherwise you can’t type the targetted characters. I ended up copy/pasting.
After about two years crafting this setup, breaking it many times, and recovering with no problem, it seems it’s becoming stable enough for me to rely on it for long term. So it just make sense to also put a proper README on it, for any other interested soul out there.
This could be useful in settings where using a Mac is mandatory by client.