Atemu

Public Key

npub1awnpt9auss5j9erjcuyqh9w3u3ykjec9ur7rz4xjs4pp63uexlys7z9yyw

NIP-05 Address

Profile Code

nprofile1qqswhfs4j77gg2fzu3evwzqtjhg7gjtfvuz7plp32nfg2ssag7vn0jgprpmhxue69uhhyetvv9ujumt0d4hhxarj9ecxjmntqyt8wumn8ghj7un9d3shjtnswf5k6ctv9ehx2aqhuxc5p

Publishing to

relay.momostr.pink

relay.primal.net

offchain.pub

Author Public Key

npub1awnpt9auss5j9erjcuyqh9w3u3ykjec9ur7rz4xjs4pp63uexlys7z9yyw

Show more details

Last Notes

2026-05-05 18:06:30 UTC

OH: MRMDB
#MRMCD #DB

2026-04-25 12:58:36 UTC

- reply

Und die zieht schenbar einen echten Zug‽ O.o
Wenn Museumsbahnen im normalen Verkehr noch einen Einsatzweck haben, sagt das einiges über die Bahninfrastruktur aus..

2025-12-27 01:38:11 UTC

Zahnpasta?
GNU/Elmex
#elmex #gnu #emacs

2025-12-15 17:36:44 UTC

If anyone has a #39c3 ticket they no longer need, I'm looking for one and will pay full price.
It'd be my first congress and it'd mean a lot to me :)
:boost_ok:

2025-12-04 17:43:49 UTC

- reply

Ja, zfs und bcachefs können das auch. Das ist allerdings praktisch exklusiv ein Feature von modernen copy-on-write Dateisystemen.

2025-12-04 09:18:22 UTC

- reply

btrfs verifiziert damit, dass alle Daten, die es auf die Platte geschrieben hat, auch immernoch exakt diese Daten sind und sich nicht verändert haben.
Somit lässt sich korruption durch Hardware frühzeitig erkennen und ggf. betroffene Daten reparieren.

2025-08-16 08:27:32 UTC

- reply

That sounds horribly inefficeient and unperfomant; that might be why.
The only sensible place to do this is the filesystem itself, so it's only natural that we saw it there first and it turns out to have significant costs even there.
Another reason might be that those who had the resources to implement that (enterprise) bought drives that were specifically certified to work with their servers' controllers. There is actually a point to that besides vendor lock-in milking.

2025-05-05 14:39:08 UTC

In @npub1z7x…q5jf's new video he mentioned how trams last a lot longer than busses do which prompted me to think.
In #Darmstadt, we still have a decent amount of #ST12 trams from *1990* in regular operation.
They're missing some features of newer models but they're not *that* far behind either. I even find them to be superior in some regards such as seat comfort; they're actually generally still quite nice.
I'd shudder to think of riding a bus from 1990 if those still existed.
#tram #lrt #bus

2025-03-17 18:45:48 UTC

#TIL #systemd can somehow restart itself in a running system and #NixOS does it automatically whenever necessary.
This must have happened dozens of times already and I never noticed because it's entirely seamless.
I only found out because I was curious what exactly the #NixOS activation script means when it says "restarting systemd" as that struck me as near impossible.
I confirmed that /proc/1/exe points at the same store path that my new version of systemctl comes from; how cool is that?

2025-02-07 14:08:04 UTC

On your #Ryzen7000 iGPU (i.e. #framework): If you visit https://openstreetmap.org and use it by scrolling, zooming etc. do you sometimes get a large multi-second hangs where you cannot interact with the desktop at all?
(That's your GPU soft-recovering, you can see it afterwards in the kernel logs.)
#AMDGPU #openstreetmap

2024-12-02 20:35:28 UTC

- reply

How exactly would you like a directory to be hashed? The hash of its NAR representation?
Also: What for? If you just want a unique identifier, you could just coerce the path literal toString.

2024-11-10 21:02:34 UTC

- reply

It means that you haven't updated and are still running 23.11 which has been EOL for quite a while now.
You need to look at the release notes for breaking changes and then update your channel to point at a newer release.

2024-10-22 10:14:00 UTC

TIL someone is building a lightweight #Android translation layer for #freedesktop and it actually works with some apps already.
It translates Android views into native #GTK widgets. Wild.
This makes it function quite similarly to #winehq rather than #waydroid.
https://gitlab.com/android_translation_layer/android_translation_layer

2024-09-06 03:35:54 UTC

- reply

Hm, that sucks.
How cheap are you getting the drives in the ~1TB range?
Note that high-capacity hard drives can be had for 15€/TB. If you're paying more than 10€/TB, I'd say you'd be better of with purchasing one big driver rather than a bunch of small ones.
The capacity shouldn't really matter in this case though as you could do the same on small drives as long as your data is sharded in a way such that each shard is <1TB.
<1TB shards would be a requirement in either case though as I see it, right?

2024-09-06 03:26:30 UTC

- reply

Sounds fragile, overcomplicated and inefficient tbh.
What speaks against buying a few large hard drives (18-22TB) and using #Borgbackup to back up your data into a regular old filesystem on those?

2024-09-03 18:23:28 UTC

- reply

Yes, hate is never productive.
The current style is imperfect at best. If you have constructive criticism that could improve it, you can provide it at https://github.com/NixOS/nixfmt/issues but please try to understand the constraints and design principles first.

2024-08-16 22:04:48 UTC

- reply

Disclaimer: IANAL.
I find that to be a quite literal interpretation of the AGPL section 13. I see two arguments against this line of reasoning:
1. If your software is not actually deployed, no user is interacting with it (note the progressive form of "interacting" in the license text). In order to offer 0 users an opportunity to receive the source code, nothing needs to be done.
2. If your software is not actively deployed, it does not support network interaction, thereby triggering the explicit exception.
Any half-decent court should also recognise that the intention of the AGPL is not to impose inconsequential requirements upon people who modify the source code in order to contribute but consequential requirements onto those wishing to run it as a service to actual users who could desire to receive the source code.
I also don't see how it'd be legit to proxy away the AGPL. That's about as legit as claiming that your only user is the router between you and your actual users because it exclusively handles the packets and therefore the router is the only user.
Both are infrastructure components of a computer network which do not significantly alter the function of the service whose function they facilitate. I don't think any court would allow that argument to stand.