Mobile security, proof-of-concept.

[David Edmundson]

> You were right, this is indeed pretty neat.

Glad to hear it.

It's certainly where Linux is heading. I've been looking into these
quite heavily purely from the POV of resource constraints and tracking
rather than security. There's so many other cool possibilities: I
especially want network namespaces to track network use per-app.

> I'm looking at the Anglefish on my pinephone that is running just
> fine in this jail.
>
> > My initial reaction is that you're absolutely right that given
> > we control the stack on plasma-mobile and expectations are
> > different we should have a dedicated app launcher that boxes
> > things a bit more than it does now.
>
> My thinking goes that apps should be isolated so they can't kill
> other apps, or even see other apps, same with IPC and they should
> not be able to read your homedir etc.

You are absolutely right that apps should not be able to kill other
apps (pid namespaces will solve that), but we still need kwin to, and
it's about finding ways to not need to add elevation to kwin which we
were first looking at; which I felt would become a continuing can of
worms.

> Most of these steps are pretty trivial, its just calling libc
> methods.
> I understand the idea of "just wrap bubblewrap", but that feels
> like we are giving much more credit to the app then it deserves.
> The amount of LOC I need to duplicate the basics is just a
> handful.

Yeah bubblewrap is 90% just doing argument parsing. I wouldn't object
to our own version.

David