Re: microkernels

[Olaf Buddenhagen]

Hi,

On Fri, Dec 19, 2014 at 07:25:34PM -0500, Kete wrote:
> On 2014-12-17 5:43, Olaf Buddenhagen wrote:
> >On Mon, Dec 15, 2014 at 02:43:20PM -0500, Kete wrote:

> >>>>Smartphone users would benefit from this x15
> >>* much less code taking up crucial space on the device
> >
> >Microkernel architectures actually need more code for the same
> >functionality than monolithical ones.
> 
> Overall, does it still result in about 1% of the size?
> 
> >>100,000 lines of code than 14,000,000.

The microkernel itself is much smaller of course. (Though total LOC is
not a good measure, as the bulk of it is for drivers, of which only a
tiny fraction is used on any particular system...)

However what actually matters on resource-constrained systems is the
total size of the code implementing the system functionality. Moving
part of the functionality into separate address spaces (out of the
kernel) doesn't make the code smaller. It actually makes it somewhat
larger, because of the overhead of isolation.

The overhead is hopefully small enough not to pose a serious problem on
any system large enough to have an MMU (microkernels are rather
pointless on systems without address spaces) -- but at any rate, a
microkernel-based system offering the same functionality is *not* at an
advantage regarding size.

-antrik-