[CLUE-Talk] gnu-hurd article

[Dennis J Perkins]

Jeffery Cann wrote:
> On Sunday 10 November 2002 09:59 am, Dennis J Perkins wrote:
> 
> 
>>Hurd does have some some interesting ideas but I don't know how feasible
>>they are.  I have wondered if Hurd is best suited to cluster systems.
>>Hurd will test these ideas and we'll see how well they work.
> 
> 
> The biggest difference is the architecture.  The GNU Hurd kernel is a 
> microkernel.  (Windows NT/2000/XP is also a microkernel).  The Linux kernel 
> is a monolithic kernel.
> 
> There have been endless debates on the merits of each type of architecture.  
> But, if you want to read more about GNU Hurd:
> 
>  + http://www.gnu.org/software/hurd/hurd.html
> 
> Jeff

The microkernel requires careful design because it uses message passing. 
  Even proper passing is not as efficient as a monolithic kernel due to 
increased overhead.  The question is whether that inefficiency is 
significant.  And whether monolithic kernels are truly obsolete.

I think the long delays in Hurd development lie more in the modules 
built around the microkernel and not the microkernel itself.  Hurd's 
complexity is due to the capabilities in those modules.  In theory you 
should be able to drop in a new module or a replacement module without 
affecting the rest of the system, possibly without rebooting the system.