arihi :bocchi_arch: :naima_padoru: :blobcatchristmastree: :ibrs2: :ene: :adorizz: :ruby

i feel like a sustainable way out of supply chain attacks would be a mix of sandboxing everything and having a web-of-trust dependency vetting tool also eliminating generative models, but that's obvi...

arihi :bocchi_arch: :naima_padoru: :blobcatchristmastree: :ibrs2: :ene: :adorizz: :ruby_happy: @arihi :bocchi_arch: :naima_padoru: :blobcatchristmastree: :ibrs2: :ene: :adorizz: :ruby_happy: 1780841665

@nprofile1q... many sandboxing solutions i've seen rely on asking the program nicely to stay in a sandbox and/or adding massive overhead. i think for any sanboxing solution to work, the program must also additionally not know that it's sandboxed which as far as i know is kinda hard to do.

for example on android (and i believe also in flatpak) a piece of software can easily usurp a full filesystem permission by refusing to work without it, because it explicitly knows if a permission wasn't given. even if it didn't have the explicit permission is can fingerprint the environment, so it's also probably desirable to allow the user to actively fake the environment that the software runs in.