Comment by rekado
1 year ago
Well, there's nobody blocking the work. If not I then surely somebody would personally have to do it. And it's akin to building a completely separate distribution on top of a different foundation, so we could only superficially reuse existing infrastructure.
I'm writing my comments in the first person, because I have actually made the effort to investigate this in the past, more than once.
This is precious little to do with some kind of abstract purity. Hell, I've packaged Tensorflow and CUDA crap, which is as far removed from purity as it gets.
It gets a little tiring to read about values that are projected onto Guix, that I can't find in my own work.
> that I can't find in my own work
But you're not the spokesperson for Guix, are you?
A quick search of GUIX mailing lists turned up this thread from 2017.
From [Chris Marusich](https://narkive.com/BnGNBXUh.7):
> I don't want to use Guix on macOS to package, promote, or make it easy to use non-free software.
Interestingly, that entire thread is a good faith exploration of what it would take to run Guix on mac at the time, but they conclude that there's no Free Software way to bootstrap the entire process, and rule it out. They consider toolchains, compilers, emulators, VMs, Docker, etc. but not starting with XCode.
Which is fine! Guix simply has different values and makes different choices.
You repeatedly say "can't", but given how Guix shares DNA with nix, which has done it on a mac, I'm a bit dubious about any technical infeasibility.
> But you're not the spokesperson for Guix, are you?
I used to be co-maintainer and I'm working close to full time on Guix. I also happen to have worked on Guix on Mac before; in fact I'm involved in the very mailing list thread you dug up.
We are familiar with the way Nix does it on Mac, but I won't repeat my assessment here. Doing it with XCode and adding an arbitrary cut to the graph would get us to something that resembles Guix only in uninteresting ways.