Sources are shared, sharing compile-time artefacts is done within workspaces.

The only time I ever interacted with python packaging was when packaging for nixos. And I can tell you that the whole ecosystem is nuts. You have like ten package managers each with thirty different ways to do things, none of which specify dependencies in a way that can be resolved without manual input because y’all have such glorious ideas as implementing the same interface in different packages and giving each the same name and such. Oh and don’t get me started on setup.py making http requests.

cached copies of crates that you downloaded

Meh, what else is it supposed to do, delete sources all the time? Then people with slow connections will complain.

Also size-wise that’s actually not even much (though they could take the care to compress it), what actually takes up space with rust is compile artifacts, per workspace. Have you heard of kondo?

Maybe there’s something I’m missing here

SFP direct attach, you don’t need a switch or transcievers, only two QSFP-DD ports and a cable. Also this is a thought exercise not a budget meeting. Start out with “We have this dual socket EPYC system here with full 12TB memory and need to double that”. You have *rolls dice* 104 free PCIe5 lanes, go.

Depends on your workload and access pattern.

…I’m saying can be faster. Not is faster.

Show me an SSD with 50GB/s, it’d need a PCIe6x8 or PCIe5x16 connection. By the time you RAID your swap you should really be eyeing that SFP+ port. Or muse about PCIe cards with RAM on them.

Speaking of: You can swap to VRAM.

Both SD 1.5 and SDXL run on 4g cards, you really want fp16 though.

In principle it should be possible to get decentish performance out of e.g. an RX480 by using the (forced) 32-bit precision to do bigger winograd convolutions (severely reducing the number of fmas needed) but don’t expect AMD to write kernels for that, ROCm is barely working on mid range cards in the first place.

Meanwhile, I actually ended up doubling my swap because 16G RAM are kinda borderline to merge SDXL models. OOM might kick in, it might not, and in any case your system is going to lock without earlyoom.

A SATA SSD has ballpark 500MB/s, a 10g ethernet link 1250MB/s. Which means that it can indeed be faster to swap to the RAM of another box on the LAN that to your local SSD.

A Crucial P5 has a bit over 3GB/s but then there’s 25g ethernet. Let’s not speak of 400g direct attach.

When the app launches, first it checks for the remove ads purchase, and if it is not present, it will initialize the ads sdk.

UX-wise you might want to add a screen full of blurb before initialising the sdk. “The free version is ad-supported, google might ask you for tracking permissions” kind of thing.

Absence of static typing is what’s wrong with nix.

That might just be tweaking default settings, KDE is ridiculously flexible. They definitely didn’t dig very deep into it or gamescope wouldn’t be a thing.

In the shame corner wearing a dunce cap for copying NixOS without making the language situation any better (like, say, static typing).

You could reasonably use C/Rust and C/Haskell to say that you have mastered the fields of unsafe and GHC.Prim.

C#, though? C#/Java would make sense given that they’re the same language with slightly different Syntax.

It’s to the AGPL what the GPL is to LGPL: More viral. Calling it “not open source” when using it would require the likes of Amazon to open up their complete stack is kinda hare-brained.

It could reasonably be argued that it’s too viral and gods know people have been doing that with the GPL but OSI’s argument “enforces additional usage restrictions” is… well, then any copyleft is a usage restriction if you want to be consistent.

The other cool kid on the block right now is bevy. It’s less of an engine for people who just want something to write their game in, though, but more of a framework for people looking to write their own engine. There’s practically nothing you can’t customise or replace in that thing, it’s built to be both flexible and performant.

There’s another way to see it: If a project has more than three contributors it’s not doing one thing, and one thing well, and should be split up.

“fuck you, build it the way we feel it should be built”

Not really sufficient on its own, you have to fit the organisation and communication structure to the code. That even applies to small teams, even to having a single coding partner, heck it even applies to a solo project you have to get the communication between your idiot and genius self right.

Indeed. Make sure to start it with hx --tutor the first time around so you know how to quit :)

And no matter what you do when giving it a try do it in a time and place where you can go at least a week without vi as the command grammar is close yet different enough to completely confuse your muscle memory, you don’t want to mix them up (helix uses a strict selection-action command set so you get ‘wd’ instead of ‘dw’ and stuff).

I mean it does support LSP, natively, I found that ultimately that’s all the plugins I really need. It working out of the box and not requiring megabytes of configuration files is one of its great strengths.

If all you need is some customisation it’s perfectly possible to write custom commands that execute sequences of commands. Including calling out to the shell and piping to and from external programs. Strictly static sequences though unlike the abomination that is vimscript they’re not making keybindings a scripting language…

You should really convert to helixism, the latest messianic update to the cult of vi.