Martin Schwaighofer

Maybe they'll add an s to the end for each subsequent release.

They could even rebrand gpt-2 and gpt-1 as os and o.

You can use fetchgit to fetch the repo and lib.fileset to do the filtering.

You're right. I was looking at this kind of in the wrong way. 😅

Oh, you probably mean just taking snapshots, not rolling back to them? 😅

It sounds like that would lead to having inconsistent data / a 'split view' of your data whenever a rollback happens.

Is this useful in some applications anyways?

They are called Genki Instruments, the product the showed at the time is a Ring called Wave you can use to control stuff.

YouTube video by NixCon NixCon2024 Deterministic Firmware with Nix

Not 100 % what you asked I think, but Olí had a great talk at NixCon 2024 about how the firmware of the product his company builds for live music performance is built with nix:
youtu.be/Nfn_srkKans

Virtual Boy 2 😄

Schiff 🚢

Though that's not the only cool kind of verification I'd like to implement. For now I'm still working on the basics, and I'll work my way up to the really crazy parts. 🤪😅

I started implementing a policy engine for trust model verification with the datafrog library now, so when someone wants to trust key A and key B only when they agree (reproducibility) that will be possible. 🐸

The GitHub UI showing a fork of the snix repository under my git username: mschwaig. The description reads: a development fork of https://git.snix.dev/snix/snix for laut The license is MIT, and there are no stars or forks of it. Nothing else is visible besides a few buttons to navigate to Branches, Tags and Activity for the repo.

Actually I even forked snix, so that I could port my upstream placeholder implementation to Rust and put it there, because that made sense to me. 😅👍

The part of the GitHub UI that shows which portion of a repo is implemented in which language. It shows 78.2 % Python, 14.2 % Nix and 7.6 % Rust.

I've started adding some rust code to laut. I even depend on sinx, the nix implementation written in rust, for nix32 encoding support.

I get it. I think the middle ground of using Nix on another distro is probably nice for a lot of people who can't afford to chase after all of the small things that need effort to get working on NixOS.

Why did you switch back? 😊

Hi Wallfacer! 😊

Whenever my computer acts up I prepare a Computer Ant with a little expedition pack (headtorch, screwdriver, sandwiches, etc) and drop it into the USB port to go find the problem. It's never worked, not even once, but it's cute and distracting and that's what really matters

That's a bit scary.

With ZFS I just kept a 1 GB reserved partition around to deal with how it behaves when it runs out of disk space.

That was only necessary to give the nix db enough breathing room to successfully do GC.

YouTube video by martin_builds_stuff laut and the supply chain security best practices checklist

I ended up actually making a video about this now. If you are interested in Nix and Supply Chain Security, check it out:

youtu.be/lqH2lVe8Isc

EDIT: changed link to fix stereo audio

Thanks, it's very nice of you to take the time to share your thoughts on this. 😊

Getting to 5 points probably isn't that easy actually, especially if you're not using Nix to do it. 😅

Let me know if you have any questions about laut.

Thanks. 😁

Thanks to the team at KTH in Stockholm, for your hard work on organizing this wonderful event.

It is great that you are giving the supply chain security community this stage, which i think it desperately needs.

In any case, when I get back to the studio at JKU I will try to make a short video about those individual checklist items, and not only if and how laut addresses them as part of its own development, buy also in terms of the vision that I have for what the project aims to provide to users eventually.

Maybe it also shows that I am willing to go to great lengths in order to advertise my own project that I deeply believe in.

Or it shows how most supply chain security professionals actually use a measured approach in terms of what they set up for their personal projects.

I don't know. 😅

I think it shows how hard it is to have a spotlight put onto what exactly you are doing in terms of supply chain security, especially in a room full of experts, even if you are yourself an expert, and when you have not checked all of the possible boxes yet, in terms of what others might expect.

Me, a male human, in my bodily form, sitting in one of the window seats of an airliner, wearing a KTH hat and holding my filled-out best practices in software supply chain security self-assessment sheet, obscuring my lower face and chest.

I won this nice hat today, for best practices in software supply chain security, for how I work on laut, at the Workshop on Supply Chain Security hosted by the CHAINS project at KTH. 🥳

I checked 5 out of 15 boxes on their self-assessment, and basically won by default because nobody else entered.

I think a lot of people and companies are getting frustrated with issues that stem from such limitations in these other tools and the huge effort required to manage them in various contexts, and are giving Nix a shot, in spite of its rough edges.

If you have two things that are packaged in Nix, and you want to build or run a third thing that depends on both, conceptually that's easy to do in Nix.

Nix makes things composable like that, tools like docker can't do that.
Instead they force you to introduce, and then break a layer of isolation.

I'm still scared of btrfs, because it was bad in 2018 or even earlier.

Like the basic example people used to show off ZFS, where they use two small files as mirrored disks and have ZFS catch and fix it would not get caught in btrfs. I think.

So, ... it better now and I can use it on my laptop? 😅

Nobody:

UDP:
sessions never meant anything to me

Thanks, you're right. The checkout action is real quick though, so I have no idea about the impact in those terms. 😅