Re: Challenge: Find potential use cases for non-trivial confinement

[olafBuddenhagen]

Hi,

On Mon, May 01, 2006 at 05:42:21AM +0200, Pierre THIERRY wrote:

> In a software development course about reverse-engineering, the
> students are given the ability to execute a program written by their
> teacher, and have to write themselve a program that, taken as a black
> box, gives identical output as the master program with the same input.
> The students must have no way to debug the program, as the course if
> focused on I/O analysis, not disassembly or source code analysis.
> 
> As the program needs significant CPU usage and the teacher has not
> enough CPU quota on the faculty's system to give it away to it's many
> students, the program cannot run as an advertised service in the
> teacher's session, but instead has to be run with CPU usage of each
> student.
> 
> For the same reason applied to the faculty or each student, the
> program cannot run as an advertised service with CPU quota given
> especially by the faculty or taken from the students collectively.

This precise use case probably should be considered valid.

However, it is extremely unlikely that this situation actually occurs in
practice, *and* poses a serious problem. (Not least because if an
algorithm is so complicated that computational power is a problem, it's
hardly a candidate for black-box reverse engineering...)

I guess it's obvious that it's absolutely impossible to create a system
that caters to every possible obscure use case. Some prioritization is
necessary, and you can hardly argue that your example has any priority
in practice.

So, the short answer is: Yes, this is something that Marcus' proposed
design doesn't offer a technical solution for. But no, it is not
something we can consider a serious problem that should dictate the
system design.

-antrik-