I get that calling command line tools is a bit clunky, but python is always my go-to when shell scripting gets too painful

Not sure I like their definition of declarative. I’d instead say that a config is “declarative” if the result of applying that configuration is independent of the current state of the system.

MacOS

I had a go at using guix as a package manager on top of an existing distro (first an immutable fedora, which went terribly, then OpenSUSE). Gave up for a few reasons:

• As mentioned in the article, guix pull is sloow.
• Packages were very out of date, even Emacs. If I understand correctly, 30.1 was only added last month, despite having been available since February. I get that this isn’t the longest wait, but for the piece of software you can expect most guix users to be running, it doesn’t bode well.
• The project I was interested in trying out (Gypsum) had a completely broken manifest. Seems like it worked on the dev’s machine though, which made me concerned about how well guix profiles actually isolate Dev environments. This was probably an error on the dev’s part, but I’d argue such errors should be hard to make by design.

All in all I love the idea of guix, but I think it needs a bigger community behind it. Of course I’m part of the problem by walking away, but 🤷

Seems like a pretty fun language with an unfortunate amount of 90s baggage.

However, I firmly believe that trying to de-parenthesise lisp is a distraction. The main reason being that s-expressions make the beloved code=data concept very obvious.

A suitable editor makes it really easy to ignore the parens (until they’re useful, e.g. for navigation). When reading, the structure of the code is inferred from indentation and line breaks. Just like C.

Most of their quotes come from this Ono guy…

Ono, who is also a freelance mathematical consultant for Epoch AI.

Ahhh, there it is.

There’s something chilling about an expert being asked if a statement represents their opinion, them saying “no”, then the statement appearing regardless but attributed to “some experts”

If you’re able to easily migrate issues etc to a new instance, then you don’t need to worry about a particular service providers getting shitty. At which point your main concern is temporary outages.

Perhaps this is more of a concern for some projects (e.g. anything that angers Nintendo’s lawyers). But for most, I imagine that the added complexity of distributed p2p hosting would outweigh the upsides.

Not saying it’s a bad idea, in fact I like it a lot, but I can see why it’s not a high priority for most OSS devs

The project’s official repo should probably exist in a single location so that there is an authoritative version. At that point p2p is only necessary if traffic for the source code is getting too expensive for the project.

Personally I think the source hut model is closest to the ideal set up for OSS projects. Though I use Codeberg for my personal stuff because I’m cheap and lazy

I think it means client-server basically. You can host a server in “the cloud” then access a frontend to it via your browser.

Might also mean it has features relevant to debugging/deploying cloud services.

Cloud is often a BS marketing word, but I’m sure there’s ways to make it justifiable in this case. (Not that any of us has to like these features. I for once can’t stand the idea of having my editor run inside a browser…)

Keep the files in a dedicated torrents folder then make symlinks to where you actually want them?

Are there any new features in particular you’re hoping for?

For me, those two are the only things I can remember thinking it would be nice to have. Q