I’m certainly not recommending snaps, but, it is important to acknowledge the problem they’re trying to solve. “The debian model” means using years-old versions of everything, having a single set of dependency versions every program must share, and giving every package’s control scripts root access while you install it. This paradigm made sense when it was developed 25 years ago but it is far from ideal today.
i still ♥ Debian but there are tons of things I need to use which I can only get from somewhere else, so, “the Debian model” for me nowadays means a stable base system and then lots of software from other distributors (sometimes flatpak or appimage, but also a lot of podman containers of various distros).
What I am almost never willing to do is use 3rd party entries in my apt sources.list file on an actual host system (though I do in containers when necessary) - down that path lies madness.
yeah, I am aware, and I do actually think the xdg portal stuff is generally a good idea for a lot of programs… but the way it works right now sacrifices a lot of usability and doesn’t gain much security.
passing files given as commandline arguments seems like an easy problem to solve, but the linked file situation with SVG is much harder (probably requires a whole new flow for xdg portals where a program can request access to a bunch of files and prompt the user once to allow access to all of them). in the absence of any solution, imo it is silly that they’re shipping inkscape as a snap with strict confinement today.
I’m unsurprised to see lots of good reasons here why not to use them already, and none for why anyone does :)
I imagine the vast majority of snap users are using them only because Ubuntu ships a few things (like firefox) as snaps by default now.
I tried the Inkscape snap recently on an ubuntu system where i needed the latest release, and found that due to its sandboxing security theater (last I heard it is still not difficult break out…) it is impossible to open files from the commandline. And, even worse, when you use the Open command from File menu, it just passes the one file you selected in to the sandbox, so, when you open a file which has references to other files (which is not uncommon with SVG) it is not able to load them! So, I ended up using Inkscape’s AppImage instead.
https://www.redhat.com/sysadmin/podman-docker-compose says root is required but https://brandonrozek.com/blog/rootless-docker-compose-podman/ (a year later) says it isn’t
in what way(s) specifically do you think he objects to the unix philosophy?
have you read his rebuttal to that claim (point #10 here)?
(disclaimer: i am using systemd on some, but not all, of my gnu/linux systems today… and after years of finding it irritating I am actually coming around to appreciate it.)
These instances are not only slow, they also sell their user’s information.
I’ve long wondered about this regarding alternate frontends for big sites in general, but especially about the youtube frontends like piped and invidious which must require substantial resources to operate (though sometimes they’re serving the video data itself directly from google, sometimes they’re actually proxying it too)… but, I’ve never seen any evidence of it. Have you?
In any case, many of the popular instances are running on Microsoft or Oracle clouds, and/or Clownflare, so, from a privacy standpoint it isn’t really much different than Google. I use them occasionally in Tor Browser just because they’re less likely to be blocking a Tor exit than youtube proper is.
these people should be ashamed of themselves and scientific american should be ashamed for publishing them. writing that they asked GPT-3 for consent is especially embarrassing.
they are contributing to this problem: https://lemmy.ml/post/343852
fwiw some other people actually did this years ago and did get at least one neural-network-authored paper past “peer review” and published; this says a lot more about the peer review process than it says about neural networks.
( https://nitter.net/cajundiscordian/status/1536503474308907010 )
The English language might not have a “single word” for the feeling I’m getting as I read more about this, but German does: fremdscham. This is actually a really sad story.
The current link in this post goes to a year-old story about the online translation feature… here is the same site’s coverage of this week’s news - which is that there is now offline translation support: https://www.ghacks.net/2022/05/30/firefox-translations-firefoxs-offline-translate-feature-is-making-progress/ (i assume this is what OP actually meant to post). (edit: OP fixed the post’s link)
Here is a web page that loads their wasm translation engline and does the actual translation offline (and it does work in the stable release of Firefox). It’s irritating that the extension still requires a nightly firefox build, as I’d like to use it in my daily browsing but I don’t want to use nightly all the time.
There are a small number of apps that have legitimate reasons for background location access, like OsmAnd which is very nice for making GPX tracks (in an offline, privacy-respecting way). But yeah “foreground location” and “background location” should be different permissions, and really, why should that app even run in the background?
(note: OsmAnd should be installed from f-droid to get the unrestricted free software version; the version in google play hilariously requires you to pay for the ability to download more than a few maps 🤣 )
alternative automated plagiarism engine?