[coyotos-dev] the GNU system design

Valerio Bellizzomi devbox at selnet.org
Fri Nov 3 12:42:59 CST 2006

On 01/11/2006, at 10.44, kragen at pobox.com wrote:

>I was just rereading http://www.gnu.org/philosophy/stallman-kth.html
>and I guess I must have not read it since before I heard about EROS
>(perhaps in 1998?) because it has this bit that I didn't remember:
>    Now, for the kernel I am planning to use a system called TRIX (it
>    doesn't stand for anything that I know of) which was developed as
>    a research project at MIT. This system is based on Remote
>    Procedure Call. Thus programs are called domains. Each domain is a
>    address space and various capabilities, and a capability is none
>    other than the ability to call a domain. Any domain can create
>    ``capability ports'' to call it, and then it can pass these ports
>    to other domains, and there is no difference between calling the
>    system and calling another user domain. In fact you can't tell
>    which you have. Thus it is very easy to have devices implemented
>    by other user programs. A file system could be implemented by a
>    user program, transparently. It's also transparent to communicate
>    across networks. You think that you're directly calling another
>    domain, but really you're calling the network server domain. It
>    takes the information that you gave in the call, and passes this
>    over the network to another server program which then calls the
>    domain that you're trying to talk to. But you and that other
>    domain see this as happening invisibly.

I believe the concept of "process ports" is borrowed from the Mach
microkernel, btw a Mach paper is referenced.

>    The TRIX kernel runs, and it has a certain limited amount of UNIX
>    compatibility, but it needs a lot more. Currently it has a file
>    system that uses the same structure on disk as the ancient UNIX
>    file system does. This made it easier to debug the thing, because
>    they could set up the files with UNIX, and then they could run
>    TRIX, but that file system doesn't have any of the features that I
>    believe are necessary.
>This was on 1986-10-30, so I guess we can backdate our efforts to
>supersede Unix with a free-software capability-secure system by
>another decade.  Maybe we'll make some progress one of these days.
>Kragen Javier Sitaker in Caracas  kragen at pobox.com
>coyotos-dev mailing list
>coyotos-dev at smtp.coyotos.org

More information about the coyotos-dev mailing list