Learned a bit o #python #gnupg bindings to automate tasks, as gpg command line can be sometimes a bit confusing and ambiguous.

E.g. this is how to purge revoked keys after gpg --refresh-keys:

#!/usr/bin/env python3

import gnupg

if __name__ == "__main__":
    gpg = gnupg.GPG()
    keys = gpg.list_keys(True)
    for k in keys:
        info_db = k['subkey_info']
        for subk in k['subkeys']:
            info = info_db[subk[0]]
            if info['trust'] == 'r':
                fp = subk[2]
                print(f"{fp} {gpg.delete_keys(fp, expect_passphrase=False, exclamation_mode=True)}")

Definitely worth of trouble because I’m super-talented on making destructive mistakes with command-line arguments :-)

have to admit that i’ve missed this before but better late than never: https://libguestfs.org/. super useful #qemu

In addition to looking how #Ubuntu #TPM2 boots I'm also somewhat interested on how e.g. #Fedora and #OpenSUSE are approaching this, or will do so in future. I have OpenSUSE in my main desktop but it is still using traditional passphrase style...

IMHO best way to address AI plagiarism would be to educate the staff on the topic.

E.g. just with human eye and simple comparison you can often spot snippets of wikipedia articles.

Also, models always lag behind so by adding as up to date stuff as possible to assignments the applicability of AI will decrease by factors.

Generally speaking AI risks should be dealt same way as vulnerabilities and malware, i.e. you should think it as a process rather than a something ending up with some final solution.

Just looking at what #Ubuntu installation stores #NVRAM of #TPM chip when installed with #TPM2 sealed #encrypted boot:

$ sudo tpm2_getcap handles-persistent
- 0x81000001
- 0x81010001

So: I guess the 2nd key, which has a policy bind, is what is “the fast path” with #PCR unsealing, and the first key is “the slow path” unsealed with the value given by snap recovery --show-keys.

Is this how it is laid out?

heapless is by far my favorite #Rust #crate as it fixes things that annoy me the most in the current standard library. I.e. it enforces explicit checks. #rustlang

I've found my first useful application for 'deep" AI: a plain spell check #aspell style. I underline not to produce to the text content itself but just correct spelling and comma mistakes.

IMHO, it is pretty good application for working in open source project such as #Linux #kernel because it improves fairness and equality between native and non-native #English speakers.

You have to remember that #English is one of the tedious #language's if you analyze it objectively. It has vast amount of corner cases and exceptions. It is dead easy only because at least in #Finland you hear throughout your life by playing video games and watching TV series (we do not dub) :-)

Still, in written English e.g. comma rules are so strange that I doubt whether I will ever really can master them....

Also I have a minor #dyslexia which makes me read sentences sometimes incorrectly in any language. This can be often seen at LKML but what can you do :-)

Going to need slightly bigger truth table...

Jori Hulkkonen - Unelmaelämää koneiden keskellä [artikkeli & video] (20.3. 2024).

"30 vuotta musiikkia tehnyt Hulkkonen kokee, että musiikkialan suuri teknologiamurros on hyödyttänyt häntä."

Video: https://areena.yle.fi/1-66455452

#JoriHulkkonen #techno #house #deephouse #Kemi #Turku #Finland

https://yle.fi/a/74-20078272

#OnThisDay, 21 Mar 1945, Hannie Schaft, an active member of the Dutch resistance known as "the girl with the red hair", is arrested at a German checkpoint in Haarlem.

She is later executed, allegedly saying "I shoot better" after the first attempt to shoot her missed.

#WomensHistoryMonth #WomenInHistory #History #WW2 #DutchHistory #Histodons

#OpenPGP #PIN https://unix.stackexchange.com/a/772382/154226

For *personal* cloud storage, I think that pay-for-quota should be the new status quo de-facto. I think bits and volts are comparable enough to have matching pricing model. Huge number of TB's or unlimited quota do not make much sense from sustainability point of view.

With similar pricing model to electricity, e.g. optional carbon compensation would be trivial to implement.

I realized that I've paid for few years 2 TB Dropbox with 100GB quota at max so it makes at least for me a lot of sense to move this of cloud. There's a few pretty decent S3-compatible options available for personal use, which is also a factor more robust than using Dropbox with its awesome apps (of which I've ever used none) :-)

Paying less and having full control of the storage works for me.

#cloud #file #storage

superb :-) i should order a new encoder because old is flakky.

https://www.youtube.com/watch?v=fFj8pUaCQU4

#polyend #tracker

Phew, was definitely worth of trouble to wipe all the IM GUI crap and take time to setup #bitlbee with gateways for Signal, Telegram etc.

Be it separate GUI apps or within a browser tabs they constantly shift context from what you are doing and getting back in phase takes always 15-20 minutes something time.

Now I just open the tmux session when I have the time and have all the messages nicely in their own queries. You also get all the nice logging features and thus have text files of discussions that you grep later and find important bits (I do have some work related stuff going on e.g. in Signal groups). I regret that I did not do this already years ago.

The European Parliament has passed a law banning certain AI applications that threaten citizens’ rights, including untargeted scraping of faces from the internet or CCTV footage to create facial recognition databases, emotion recognition in the workplace & schools, social scoring, and predictive policing. https://www.europarl.europa.eu/news/en/press-room/20240308IPR19015/artificial-intelligence-act-meps-adopt-landmark-law

Bluesky makes conference time table changes great again i'd guess :-) exactly like old twitter which was great for that. have peek at some point what sort of api they provide.

I like what my ex-employer is doing with #x86s, nice work. it takes talent and skills to transform #microarchitecture with decades of legacy and move it forward like this.

#x86 #intel

#rclone is quite great, too bad i have not found it earlier.

It fixes the glitch that I've had with cloud storage, which works like Dropbox does. I can do "scattered" syncs with trivial scripting, i.e. when the files to be synced are scattered around the filesystem.

A great example is audio plugin presets and their associated files (e.g. samples) that can be deployed to multiple locations.

Also e.g. in Linux and macOS versions of U-he plugins there's some differences in the directory hierarchy. I can sort that out with two scripts, one for Linux and another for macOS, and call it day.

Life quality has been upgraded :-)

FLOSS folks, is there any research into using image pulls / downloads (or even, yuck, GH stars) as a metric for number of installations/active users?

(Please don't reply with lists of pro/cons/limitations/constraints etc, got plenty of those myself; I'm specifically looking for existing papers or other publications, thank you! 🙏)

#FLOSS #OpenSource #OpenSourceMetrics

I reconfigured my main desktop PC used for development with UKI kernel because it has been in recent past discussed at LKML around patches that are relevant in the areas I'm working on and I know almost nothing about it. So getting familiar I suppose :-) The basics are easy to understand, and the specs are nice and clean but really need to see how it works over time in daily use...