Re: hurd-l4 VM server

[Farid Hajji]

> > Seriously though, all that is needed for the Hurd is right
> > now Mach's VM interface. It _may_ be possible to write some
> > glue code that relies only on a (quite limited) subset of
> > uvm_extern.h to implement the [mach_]vm_*() functions.
> > But if it is not too time consuming, a 1:1 IDL interface to
> > uvm_extern.h would be convenient anyway...
> 
> Committing ourselves to the mach interface might not be the best
> thing. Rather use a uvm <-> mach glue code to get things work. In the
> future a good interface is the most important thing and UVM looks much
> more promising than the mach interface.

You've probably got me wrong here.
I fully agree that the UVM interface
is superior in every respect, and it
would be great if baux implemented it.

Implementing Mach-VM interface on top
of UVM (glue code) is only needed now,
to start porting glibc and the Hurd
servers that rely on mach_vm_*() functions.
That's all.

In the long run, it may be a good thing
to adapt glibc and the Hurd sources, to
use something more generic than Mach-VM;
this thing could then be implemented using
the UVM interface (or recursive pagers,
or whatever). Though I don't see anything
wrong with the mach_vm_*() interface; at
least for the moment. (IMHO).

Ideas like recursive pagers et. al. are
very nice, but who is willing to work on
that right now and contribute real code? ;-)

-Farid.

-- 
Farid Hajji -- Unix Systems and Network Management.
http://www.farid-hajji.net/address.html
Quoth the Raven, "Nevermore." --Edgar Allan Poe.