Hello Justin,

welcome to the Parrot project, and thanks for your email.

Am 02.04.2012 18:47, schrieb Justin L. Harper:

As a developer of a downstream project (Rakudo), I'd be very interested
to see such a project implemented.

There's no reason the child wouldn't be allowed to tighten its own
security -- it just can't loosen it. In fact that's the model that UNIX
systems use for things such as process priority (the "nice" value)

PDD 18 [1] describes the high-level goals and mechanism. A GSoC project
surely isn't enough to implement all of it, but having *something* would
be a good start.

Another good source for inspiration is prior art in linux. There is the
ptrace mechanism for intercepting system calls [2], and selinux for
capability-based access control [3].

I haven't spent too much time thinking about the security system, so
take the following with a grain of salt.
I think the best approach to getting results fast is to implement
something like ptrace, where possibly dangerous operations (memory
allocations, IO, ...) trigger a callback, and that callback can then
allow or forbid the operation. For performance reasons it might make
sense to allow or forbid some operations right away without any
callback, but I guess that's only a second step.

Once such a system is in place, you can write a capability-based system
on top of that.

[1]
http://docs.parrot.org/parrot/devel/html/docs/pdds/pdd18_security.pod...
[2] http://www.linuxjournal.com/article/6100
[3] https://en.wikipedia.org/wiki/Security-Enhanced_Linux

And I look forward to see the actual project proposal and the code :-).

Remember that the application deadline is the upcoming Friday, so don't
waste any time.

Cheers,
Moritz
_______________________________________________
http://lists.parrot.org/mailman/listinfo/parrot-dev