Hi, On Sat, Dec 08, 2007 at 01:48:36AM +0530, Shakthi Kannan wrote:
> ----- On Dec 8, 2007 12:32 AM, Shakthi Kannan <[EMAIL PROTECTED]> > wrote: | I would like get a big picture on how servers, ports, message > queues, | libraries, system calls, IPC, gnumach, drivers, hardware et. > al. are | layered. > > Is this something close? > http://shakthimaan.com/downloads/hurd/gnu-hurd-mach-system-architecture.png The Mach part looks about right, though I'm not sure the choice of components is optimal. The userspace part is totally wrong. For one, (almost?) all Hurd servers *are* translators. A translator is basically any process that can be contacted through the file system. Hurd libraries are used by translators to talk to Mach as well as to other translators. Note that communication with other translators always goes through Mach ports. (Well, theroretically communication is also possible using shared memory, where Mach is needed for the setup but not the actual communication, but this is used only in very special cases.) Also, glibc is a crucial component that is totally missing in your picture. It is used both by normal application and translators, when talking to the system servers. -antrik- _______________________________________________ Bug-hurd mailing list Bug-hurd@gnu.org http://lists.gnu.org/mailman/listinfo/bug-hurd
