Re: GNUstep Projects and Cambridge

[David Chisnall]

On 26 Jun 2013, at 12:25, Vasileios Anagnostopoulos <fithis2001@gmail.com> 
wrote:

> On 6/25/2013 12:28 PM, David Chisnall wrote:
>> On 25 Jun 2013, at 09:50, "Dr. H. Nikolaus Schaller" <hns@goldelico.com> 
>> wrote:
>> 
>>> I find this interesting from a theoretical / structureal point of view, but 
>>> why do you want to run any code in the kernel space?
>> Because we can :-)
> Isn't this experiment already done in IOKit with C++?

No.

>> 
>>> Do you expect a performance benefit or more security (I would expect less)?
>> Things interacting with the kernel subsystems directly would be faster than 
>> if they needed round trips to userspace.  Things written in higher-level 
>> languages should (in theory) be more reliable.  It's about finding 
>> additional places on the spectrum between performance and reliability.
>> 
>>> Or would any kernel component become easier to develop or test?
> I would expect a great reduction in LOCs.

Maybe, although it's not clear what the improvement just from Objective-C would 
be.  The kernel already includes an ad-hoc vtable-based object model and some C 
macros for creating classes / instances / methods.  Bridging this with an 
object model that supports duck typing is quite an interesting research 
problem...

>> We'll find out.  It's not research if you know all of the answers up-front...
> For me it would make much more sense to go with XNU and convert the IOKit to 
> Objective C too.

IOKit replaces DriverKit, which was the NeXT framework for device drivers, 
which was written in Objective-C...

David