Re: Hurd on Scheme/guile

[Leonardo Lopes Pereira]

Em Fri, 30 Dec 2005 06:00:23 +0100
pancake <sergipop@mx3.redestb.es> escreveu:

> On Thu, Dec 29, 2005 at 08:52:08PM -0300, Leonardo Lopes Pereira wrote:
> >     After that movement to revive/improve GNU Mach, I decided to think in a 
> > way to improve Hurd. After some talks on IRC I thought that a good start 
> > point would be the use of a new programming language to the detriment of C, 
> > something that turn translator code clear and that is powerful.
> >     I think that nothing better than lisp to fill those requirements. 
> > Looking for GNU projects to implement LISP, I decide that the modularity 
> > and the small size of Guile would simplify and turn that new Hurd 
> > implementation even better.
> 
> The GNU manifest talks about this. It's like a prophecy. :P
I know that GNU manifesto talk about that, but this is not the most important 
thing IMHO. ;)

> We discussed this in the Bee project, resulting the creation of the "stut". 
> (a sotware packaging manager that frontends stow and manages binary pkgsrc 
> packages).
good :)


> >     I do not want to describe everything on this mail, so if you have a 
> > question, a sugestion or a criticism, fell free to send a answer ;)
> > 
> 
> Jemarch does that some time ago, AFAIK he started implementing a trivfs 
> wrapper for the GNU/Guile.
>
I never read anything about that project, but what I wanna do is not only 
create wrappers to hurd libraries, but redesign them in Scheme/Guile. To turn 
them simpler and beaulty.

> 
> BTW, in Bee we're planning to use guile for the base system to use DMD and 
> Stut.
> 
This massive use of guile can generate another good point, that is the 
improvement of guile :)

> >     The first point that I wanna expose is that the goal of this project is 
> > not only complement the current translator interface, but create a new 
> > interface that turns possible a complete rewrite of Hurd on Scheme.
> 
> Stut will use the unionfs and it's planned to also use the trivfs for package 
> query "virtualization".
> 
> >     So It cannot be done only creating bindings of hurd libraries, it 
> > requests the creation of a new framework to create translators. Only that 
> > way we will be able to have a simply and powerful framework.
> >     To minimize the efects of the lost of performance caused by the use of 
> > an interpreter, we will need to use guile-lightning (based on GNU 
> > lightning), making the translator be compiled at runtime, so, the lost of 
> > performance will be concentrated on the time of turning the translator up.
> > 
> 
> This makes a weighter process and a hard bootup. It depends on the translator 
> task.
>
Sorry, I didn't get the relation between my texts and your text.

> > *A small roadmap
> >  - Turn possible the creation of Mach Interfaces in Scheme
> >  - Create a MiG that generates Guile code instead of C.
> 
> Would be better to fix the actual MiG. He needs more love.
I don't think so, the actual MiG only generate C code, I want something that 
generate Scheme/Guile code. Anyway, I think that using guile, we will not need 
to use a MiG. The manual generation of interfaces will be very simple.

> >  - Create a trivfs module
> 
> This is partially done.
There is a trivfs wrapper, not a trivfs module :)

> 
> >  - Discuss how others modules would be and improve trivfs module
> 
> This would be a nice task. Scheme is a nice language and the GNU system needs 
> something good like this :)
> 
> >  - Rewrite hurd ;)
> > 
> > Thanks AMS for some explanations
> > That's all.
> > 
> > 
> > _______________________________________________
> > Bug-hurd mailing list
> > Bug-hurd@gnu.org
> > http://lists.gnu.org/mailman/listinfo/bug-hurd