• 0 Posts
  • 36 Comments
Joined 10 months ago
cake
Cake day: January 19th, 2024

help-circle

  • I don’t know why that comment is collecting downvotes. They are referencing George Orwell’s “Animal Farm.”

    Context: “Animal Farm” is a story about how communism can devolve into dictatorship. In the story, the animals on a farm drive out their tyrannical drunkard farmer. They write on the barn wall: “all animals are equal” and live in communist utopia. But some animals, too, hunger for power and status. Rather than overturn the system, they undermine it by adding “…but some animals are more equal than others” to the barn wall, legitimizing a ruling class (themselves) because they are “more equal.”


  • When you have a bunch of computers networked, each of them is assigned a unique number, so when other computers send data on the wire, they can say who it is meant for (imagine each blurb of data starting out like: “yo, I’m sending these next 500 bytes for computer 0A123FBC32, here they come”).

    Now the right computer will listen, but it doesn’t know what program the data is for - is it a chunk of a file your browser is downloading? Or the email your email app wants to display? Or perhaps a join request from your buddy’s computer for the Minecraft game you’re hosting?

    So in addition to the unique number of the target computer, the data also specifies a “port number”, which tells the computer which of its running programs the data is meant for (programs ask the computer’s operating system: “if any network data arrives on port XY, give it to me”). Some ports have become standards - for example, a program that serves web pages to other computers would typically ask the operating system that any data arriving on the computer that indicates port numbers 80 and 443 should be given to it, and when a web browser wants to fetch a web page, it will send a request to the computer serving the page, defaulting to port 80 o 443.

    If you dig deeper, you’ll find that there are even more unique numbers involved and routers/firewalls let data through not only by port number but also by distinguishing between data that is the initial request to another computer’s port number and data that is an answer to an earlier seen request – and more.


  • Linux Unix since 1979: upon booting, the kernel shall run a single “init” process with unlimited permissions. Said process should be as small and simple as humanly possible and its only duty will be to spawn other, more restricted processes.

    Linux since 2010: let’s write an enormous, complex system(d) that does everything from launching processes to maintaining user login sessions to DNS caching to device mounting to running daemons and monitoring daemons. All we need to do is write flawless code with no security issues.

    Linux since 2015: We should patch unrelated packages so they send notifications to our humongous system manager whether they’re still running properly. It’s totally fine to make a bridge from a process that accepts data from outside before even logging in and our absolutely secure system manager.

    Excuse the cheap systemd trolling, yes, it is actually splitting into several, less-privileged processes, but I do consider the entire design unsound. Not least because it creates a single, large provider of connection points that becomes ever more difficult to replace or create alternatives to (similarly to web standard if only a single browser implementation existed).


  • I have a Windows VM that runs Visual Studio and a small number of developer tools so I can test my code on Windows. And another windows VM that runs Daz3D, Clip Studio Paint and the Epic Launcher (to download stuff from the Unreal Engine Marketplace).

    Sometimes I misuse either VM by creating a snapshot and installing Garmin Connect so I can update the music library on my watch :)


  • SuSE Linux (a German distribution), some niche, single CD distrubution, Debian for a while and, finally, since ~2006, Gentoo on my servers and since ~2015 Gentoo as my desktop.

    Debian and its derivatives never felt right for me. I find too many drawbacks with binary packages (non-configurable build options, therefore dependencies that can’t be disabled, relying on humans to keep ABI compatiblity, trouble integrating my own packages or unstable versions) and I just don’t like systemd.

    It’s weird, I’ve seen more than enough of those “Install Gentoo” memes, but I find it the most pleasant system to run in the long term.


  • I’m a little put off by the inconvenient command line and the mandatory bells and whistles (flathub is nice and all, but must it be baked into the main executable rather than having the package manager as an optional thing on top?).

    So far, AppImage just looks superior to me. Works without installing a runtime into my system, no need to become root and integrate an app into a system-wide managed package repository, I can just run it.





  • I’ve done this (shared 3 NTFS partition in a dual boot setup) from 2017 to the end of 2023 without issues.

    The trick was to disable “fast startup” and hibernation. Otherwise Windows happily shuts down with the file systems in an inconsistent state. It’s just a question whether one can live with that in their Windows install.


  • I’ve used the old ‘ntfs’ driver that supposedly can’t write to… write files ranging from 100,000+ small files in folders to individual 200+ GiB files on NTFS partitions. It works pretty well and I have used it for video editing (few huge files), software development (many tiny files), Unreal Engine + Unity, Linux Gaming w/Steam and more. Rock solid.

    After hearing that the ‘ntfs’ driver is supposed to be read-only, I switched to ‘ntfs3’ instead of using ‘ntfs3g’ (same code, but compiled into the kernel instead of running outside via ‘fuse’). From that point onwards, I’ve had major file system corruption nearly every day:

    • Copying files into folders suddenly made 90% of other files in the folder disappear. Could be fixed by copying about 1000 random files into the folder and deleting them, then the missing files would come back into existence.
    • Files that suddenly go bad. Can’t be written to, moved or anything. Often happened in software development when compiling my project, suddenly the intermediate build directory was bust due to undeletable files.
    • Folders that suddenly contain themselves or one of their parent folders as sub-folders.
    • Folders that contain a specific file infinity times. This way, I found out that even a harmless file manager like KDE’s Dolphin can become a behemoth that eats 100+ GiB of RAM and keeps trying to read the “list” of files in a directory without limit.

    Personally, I’ll never use ‘ntfs3’ for serious work again. But ‘ntfs3g’ is generally considered very stable, maybe my issues are specific to ‘ntfs3’ or my RAID setup (weird nested mdraid thanks to Intel) is to blame.

    My final ‘fix’ was to move everything to ext4 and buy Paragon’s $20 ext4 drivers for the dual boot Windows install. It’s only seeing any use once every 2 months. Sadly, these drivers are case sensitive even on Windows, rendering Bethesda games unplayable when installed on those partitions, for example.


  • A perfect demonstration of how Russian indoctrination works right here.

    Original reporting: A major disinfo attack against Europe being prepared by Russia is uncovered through diligent investigation and published and reported on.

    The response:

      1. divert to farmer’s dissatisfaction with several policies
      1. cast disinfo reports as underhanded attempts (by politician Russia wants gone) to arrogantly brush off farmer’s concerns (which the report never even related to)
      1. claim Macron is selling out to EU (here, have a serving of anti-EU sentiment, too)
      1. vaccinate reader against the disinfo being countered (“everyone who tells you otherwise belittles you and hates you, join us in our righteous anger”)

    Emotional framing:

    Nationalists, agricultural owner-operators, and farmers exposed to rising interest rates

    “truckloads of exported Ukranian agricultural salvage” vs. “fresh French produce”

    we’re getting an earful about how all these local yokels are hoodwinked by anti-EU Russian Propaganda

    Macron for selling out the agg sector to financial interests in Brussels

    “If you’re not in favor of (insert supposed evil acts described in lurid way), then you’re a secret spy for Putin and a traitor.”

    Result: The reader comes out the other end an angry person, outraged about the plight of farmers, outraged again at disinfo reports supposedly serving to silence them, outraged once more at a France politician selling them out to the EU, EU painted as high-and-mighty villain, automatic anger against anyone who tells them a different viewpoint ready to trigger.



  • Considering that…

    • The Republicans encouraged him multiple times to buy it
    • He quickly stopped blocking (mainly Russian) state-sponsored social manipulation campaigns
    • He allowed right wing agitators back on the platform
    • He almost immediately banned droves of journalists that weren’t blindly Pro-Russia and Pro-GOP.
    • He censored / banned all kinds of activists that pushed back against authoritarian (Russia-backed) regimes in other countries

    …I have a hunch that he also served the interests of certain political actors with the acquisition. Public town square my lower backside.


  • I agree that a lot of human behavior (on the micro as well as macro level) is just following learned patterns. On the other hand, I also think we’re far ahead - for now - in that we (can) have a meta context - a goal and an awareness of our own intent.

    For example, when we solve a math problem, we don’t just let intuitive patterns run and blurt out numbers, we know that this is a rigid, deterministic discipline that needs to be followed. We observe and guide our own thought processes.

    That requires at least a recurrent network and at higher levels, some form of self awareness. And any LLM is, when it runs (rather than being trained), completely static, feed-forward (it gets some 2000 words (or 32000+ as of GPT-4 Turbo) fed to its input synapses, each neuron layer gets to fire once and then the final neuron layer contains the likelihoods for each possible next word.)



  • Is this a case of “here, LLM trained on millions of lines of text from cold war novels, fictional alien invasions, nuclear apocalypses and the like, please assume there is a tense diplomatic situation and write the next actions taken by either party” ?

    But it’s good that the researchers made explicit what should be clear: these LLMs aren’t thinking/reasoning “AI” that is being consulted, they just serve up a remix of likely sentences that might reasonably follow the gist of the provided prior text (“context”). A corrupted hive mind of fiction authors and actions that served their ends of telling a story.

    That being said, I could imagine /some/ use if an LLM was trained/retrained on exclusively verified information describing real actions and outcomes in 20th century military history. It could serve as brainstorming aid, to point out possible actions or possible responses of the opponent which decision makers might not have thought of.


  • I’m not up-to-date with current NAS systems anymore – I’m running an older QNAP NAS (TS-453), and it has their proprietary “Container Station” which can run web applications in Docker + LXD containers. Not FOSS, though the containers very much are and can be moved to other systems.

    As an alternative, FreeNAS/TrueNAS sells NAS systems where at least the software side is FOSS. They’re quite expensive, though.

    The prices of other brands also quickly breach silly levels, but a basic 2-bay NAS is about ~$250 for QNAP, ~$200 for Synology and ~$1000 for a TrueNAS. Without hard drives.

    If you’re not interested in the data storage side, a Mini PC w/Proxmox (popular Docker/LXD container engine w/browser-based management) or even a direct install on a Raspberry PI are possible for under $100.


  • I’m running a few on my NAS:

    • Taiga to manage projects. It’s as easy and pleasant to use as Trello, but with velocity/burndown charts and the whole “agile” thing, but you can also turn parts of it on and off (per project even).

    • Trilium completely cured me of messy note-taking habits, simply by winning on the convenience side. I was firmly in the “folder tree of markdown documents” and “my Sublime Text tabs of random notes have no number” camp before.

    • I’m considering Habitica which lets you set up rewards and achievements for your real life (i.e. apply addictive reward/progress loop from video games to motivate your real self to do things). Also Wger for exercise tracking, but I’m not sure they’re the right thing for my ticket/tracking-averse self (I wish there was something that covered the whole MyFitnessPal/FitDay and the whole Polar Personal Trainer/Garmin Connect side, but FOSS and self-hosted).

    For leisure, I also run Stash (it bills itself as an organizer for your porn library, but it’s really good for any kind of clips), Jellyfin for my music and movies and currently both Mango and Kavita for books and comics.