OSNews

Of course! Right now, the Genode development is primarily geared to special-purpose applications where the low-complexity trusted computing base of a microkernel-based system is immediately beneficial. For example thin clients used in security-sensitive environments or point-of-sales devices. But with its current feature set, it could already be used as the underpinning of a limited OS in the line of Chrome OS.

Ultimately, however, the project strives towards general-purpose computing. For this reason, self-hosting Genode on Genode is one of the most important milestones for 2011.

Indeed, your initial wording sounded like OSs with a special purpose are not "real OSs", a way of thinking which is the starting point of software bloat

Edited 2010-12-24 14:50 UTC

Well... that's not really how libraries work.

I don't have the knowledge to explain why, but I do know that's the case. I hope someone with more info can help you out.

Yeah, it would be interesting to hear an explanation ... I mean what would it take to compile something like KOffice under Genode?

What would it take? You have to have some kind of support for an underlying windowing system. Video drivers, screen coordinates, bitmaps, mouse input, etc. This is not rocket science but it must be written by someone first.

Well Arora seems to work pretty well under Genode, so a lot of work is already done. I'm asking what's left?

Fortunately, the pieces you're mentioning are already there.

I haven't looked into koffice yet. One would need to take a closer look at kdelibs and their respective dependencies. These dependencies must be resolved by either porting the libraries or reimplementing their interfaces. For example, if koffice happens to rely on dbus, we would need to find a way to emulate (parts of) the dbus API on Genode.

That said, I think there are quite a few interesting engineering problems to tackle but I see no fundamental road block that would prevent such an endeavour.

There is no general answer to this question because many Qt4 applications use libraries in addition to Qt4. If an application is a pure Qt4 application (this is the case for Arora) it can be compiled for Genode without source-code modifications. For example, "porting" the text editor shown in the demo was just a matter of minutes. All it took was a little tweak in the qmake project file.

If, however, an application relies on further libraries, a simple recompilation of the application won't suffice because the needed libraries must be ported as well. The infrastructure for such a porting work is already there (called 'libports' repository) and some common libs such as freetype, libpng, zlib, and jpeg are already available. But there are no kdelibs yet, which would be needed for many popular Qt4 applications. Such a project would certainly be a good next step to go for.