My "production stuff":
- https://github.com/festivus-es/festivus - public holidays calendars for Spanish cities
- https://github.com/remote-es/remotes - companies hiring in Spain for remote positions
Usable WIPs:
- https://alexpdp7.github.io/selfhostwatch/ - track self-hosting package updates (such as YunoHost)
- https://github.com/alexpdp7/ubpkg/ - package manager for "upstream binaries"
- https://github.com/alexpdp7/termflux - Miniflux terminal client
Good news for everyone working with #LXC and #Incus containers: Rocky and Alma Linux 10 container are now available!
The official image server of the #linuxcontainers project added version 10 of #rockylinux and #almalinux at the end of August.
https://www.geekersdigest.com/rocky-alma-linux-10-images-available-for-lxc-and-incus-containers/
When developing software, I've long been jealous of system administration's "error budgets". If you are above your reliability target, you should be looser and faster. If you are below, you should be more careful and slower (address "tech debt").
This needs to be refined, but I think deadlines work to create a similar self-regulated system; management should be aware that setting deadlines or not having deadlines regulates the development process, like setting the error budget does.
whoever implemented KDE's "waggle the mouse to make it bigger" thing and decided it should have no end point of getting bigger should win a software engineering award
5️⃣0️⃣ Here's the 50th post highlighting key new features of the upcoming v258 release of systemd. #systemd258
User namespaces are weird beasts: on one hand they are supposed to be something that you can acquire without privileges, but on the other hand if you want more than a single UID mapped into them, you need multiple UIDs, and that's a resource you cannot acquire without privs.
To deal with that multiple systems have been devised.
The computing industry is weird about history. People fail to understand it in two opposite ways:
A lot of things just fail to pay attention to the fact that other people have seen and solved the same problem before. You get people not just reinventing the wheel, but ignoring the huge design space of wheels that’s been explored and deciding that cubes are ideal shapes for wheels and concrete the ideal building material. See pretty much every recent GUI for hundreds of examples.
But at the other extreme you get a weird worship of the past. Smart people in the past did something this way, therefore we must do things that way! The fact that their target audience was PhD students and their constraints included a processor that was barely faster than a pen and paper and had 128 KiB of RAM is irrelevant. Early UNIX did glob expansion in the shell not because that’s more sensible than providing a glob and option parsing API in the standard library, but because they didn’t have enough disk space or RAM to duplicate code and they didn’t have shared libraries. If you have shared libraries, the right choice is very different. Similarly, ‘everything is a text stream’ is a good idea when you have a computer that you connect to via a serial terminal because everything that the user produces or consumes is a text stream. When you have a system that has graphical displays, speakers, cameras, microphones, and network connections, it is much less sensible.
It’s easy to forget that these days we may have colleagues in the software industry for whom an iPad was their first ever computing device.
So just to remind you: The idea that software should come from a central authority is new, it’s radical, and it’s wrong.
https://guild.pmdcollab.org/@StaticR/115098597647157516
[Continuation of my previous post about QEMU-Wasm]
Also check out elfconv (AOT ELF-to-Wasm translator) by Masashi Yoshimura
More than 60 times faster than emulation.
https://yomaytk.github.io/elfconv-demo/
My daughter, who has had a degree in computer science for 25 years, posted this observation about ChatGPT on Facebook. It's the best description I've seen:
Figuring out memory usage on Linux (and likely other modern operating systems with modern applications) is not as easy as one would expect.
By asking questions around people who know everything, I've found out about systemd-cgtop, which I added to smem at:
https://github.com/alexpdp7/alexpdp7/blob/master/linux/misc.md#memory-usage-queries
AWS is easy (not). I have only managed very simple accounts for personal use. Now I wanted to set up an account with support for multiple users and environments. Taking some notes here:
https://github.com/alexpdp7/alexpdp7/blob/master/misc/aws/account_setup_notes.md
A few hours into trying out Trixie on my sandbox laptop. Quite uneventful; the biggest change is that 2022 Gnome was getting a bit long of the tooth and some extensions I wanted to try out are not available for that.
(I can wiggle my mouse now to embiggen my pointer. Yay.)
My provisioning scripts only required minor tweaks.
Now five "production" hosts to update. But that will wait a bit.
A little RSS planet-like toy I made:
https://alexpdp7.github.io/frozenplanetoid/beings.html
It's an SSG, so it can be hosted for free on GitHub Pages or less-MS-made services, or even on tildes.
It's not even close to beta right now, but details in the repo:
Jetbrains just sent me an email for an interview about developing blockchain with Rust.
What day is it? OF WHAT YEAR?
You can install Docker Desktop, or equivalent tools, on Linux, to assess how things behave on macOS (and Windows without WSL, if anyone cares).
However, I understood immediately why there are so many projects to replace Docker Desktop.
The 2025 Stack Overflow Developer Survey is out, and it's an important resource. I already mentioned[1] how I'm worried about the future of the survey; perhaps more than about the q&a site.
[1]: https://alex.femto.pub/@coder/posts/458497280777236416/
Deflating a bit about Supabase after figuring out Docker is practically a hard requirement for local development, despite Podman support claims.
(Which is a pragmatic decision, but not ideal for my tastes.)
Supabase looks nice. I didn't realize it was such a thin wrapper around things like PostgREST. So many people relying on Haskell for production. (And it also has an Erlang component.)
Doing some self-learning at work about Next.js and Supabase. About Next.js, maybe modern full stack development has figured out how to be as usable (or more) than traditional server-side rendering.
And Next.js seems to *default* to something that works on lynx.
Those are just first impressions, but who knows...
I'm presenting Knuth's LR parsing paper at Papers we Love Berlin tomorrow. Parsing nerds (and other types) come hang out. https://www.meetup.com/papers-we-love-berlin/events/307287436
I finally sat down and figured out a process to establish a virtual network over two hosts when I only have a pipe (such as SSH).
https://github.com/alexpdp7/alexpdp7/blob/master/linux/remote-networking.md
There must be a better way, but this works.
My biggest concern about the future of StackOverflow is their yearly developer survey. Large developer surveys are hugely useful.
(Apparently, the new survey came out, so I took the time to do my duty.)
Signal famously doesn't allow third party clients. Molly currently flies under the radar but their official stance is: fuck off.
If #Conversations_im were to introduce ads tomorrow you just switch to a fork and move on with your life. With Signal you do what exactly?
Also, I'm a bit short on hardware so I was overthinking things until I realized I can just install Fedora 42 to an external USB drive without wiping any machine. Works fairly well when you realize that you have to use the blue USB ports.
So with Fedora 42, you can create VMs that use Venus to expose a GPU to VMs that passes through Vulkan commands to the host graphics card. Meaning I can create VMs to run an application that requires a GPU (using Intel 12th graphics!). You have to use raw QEMU because libvirt does not support the necessary options yet.
Obviously not the best for performance (but I don't care), but you don't have to source datacenter GPUs or strange patches- standard Intel integrated graphics + Fedora packages.
The Promised LAN: https://notes.pault.ag/tpl/
Another piece in the communities, Yggdrasil, new Internet, etc. puzzle...