Re: If QNX is successful, why NOT GNU Microkernels

[Danilo Segan]

Hi Anand,

Anand Raj <anandraja2000@yahoo.com> writes:

> Hi all,
>  QNX is able to be a successful MicroKernel, but why NOT GNUMach /
>  OSKITMach / L4Mach ? So, what could be the problem with GNU

Mach is completely separate (micro)kernel from L4, and last I've
heard, there's no such thing as "L4Mach".  Also, it should be "GNU
Mach" and "OSKit Mach" (I only guess for this one, there may be some
different punctuation in "OSKit").

> Microkernels ? QNX is using GNU tools and they are successful. Where
> we are lacking ? coding style / documentation / info on creating a 
> MicroKernel ? QNX don't have partition limitation like we have with
> ext2. Even device driver is also implemented without any problem...

What is lacking is people who are skilled enough *and* have enough
time to work on Hurd.  Rare are those hackers who are paid to work on
Hurd (if any), so they have to devote some of their (limited) free
time on this task.  

Anyway, there're still a couple of motivated  hackers who help GNU
Hurd live.  If you follow relevant mailing lists for some time, you'll
learn fast who they are.

What should be clear is that the goal of GNU Hurd is not to develop a
microkernel (it does include making microkernel better suited for
Hurd), but to create a bunch of "servers" which run on top of (any)
microkernel such as GNU Mach or L4.  With such chosen design, it was
risky effort a priori, because microkernels have only entered the
scene at the time of Hurd birth. Many problems of microkernel design
were only resolved in the recent past. 

On one practical issue that you mention: ext2 filesystem translator
has recently been worked on by Ognyan Kulev, and he's very close to
fixing that limit, if it's not done already.  I guess this is the
thing that bothers people the most, as that's what people are most
complaining about.

As for why is QNX "successful", you'd probably have to ask its'
developers as to what is their secret formula.

Bear in mind that I'm not a developer of GNU Hurd, nor am I involved
with it in any way except as a part-time user, and all of these are
only my perceptions of the GNU Hurd and GNU/Hurd community.  They may
be completely false, and I'm sure someone will pop in to correct me
where I'm wrong.

Cheers,
Danilo