On Sun, Nov 23, 2008 at 10:25 AM,
<olafBuddenhagen@gmx.net> wrote:
I think Marek's work is interesting on several levels. For one, Mach was
originally created by taking a BSD kernel, throwing out the UNIX stuff,
and implementing some generic primitives (IPC, external pager interface)
instead. Doing the same with a modern BSD kernel actually seems the most
pragmatic approach to getting a modern Mach...
Hm, interesting. Do you think this can be adopted as a basis for a
full-blown strategy of modernizing the Hurd and making it more
interesting to ordinary users?
Adding the Mach functionality while leaving in the UNIX stuff in place,
also seems an interesting option: It would allow playing with Hurd
stuff, while still running a mature system at the same time. In fact, we
pondered the idea of running Hurd on Linux in the past...
There are two somewhat distinct use cases for that: For one, a complete
Hurd could be run on top of some other system, as a stable base for
experimentation, without having to care about the microkernel stuff.
The other is related to the hurdish application stuff I have been
repeatedly talking about. Having a partial Hurd environment running on
top of other systems would be very helpful here I believe: It opens the
possibily of writing hurdish applications, and yet making them available
to a wider audience, by offering a way to run them on other systems as
well.
I am deeply convinced that this would help spreading the Hurd. People
could easily test the hurdish stuff on their normal system. They could
see that these concepts really make sense -- and eventually grow a
desire to use The Real Thing, to get even more benefits.
I don't want to look stubborn, but the following question comes to my
mind: can it be guaranteed that the Hurd running on top of another
system will be exactly the Hurd running on top of Mach? My question
arises out of some obscure cygwin issues I've come across while
groping on the Internet: people would sometimes complain of the fact
that things which work okay on Linux won't work like that on Windows.
The issue you are raising here is really a generic one. People are often
debating whether it's a good thing or a bad thing to port free software
to non-free systems for example. IHMO it's decidedly a good thing:
Inertia is the main obstacle for people to migrate to another system --
they don't want to give up their familiar environment. Being able to use
free applications while staying in the familiar environment, allows
people to get familiar with the free software by and by, and the final
move to an entirely free system will be considerably less painful.
Note that the initial success of the GNU software stemmed mostly from
the fact that GNU programs were running on all kinds of UNIX systems...
Hm, you are definitely right. The only thing that troubles me is the
question I've just asked above.
I briefly covered some of these aspects in:
http://tri-ceps.blogspot.com/2005/09/welcome-to-hell.html