Path: utzoo!utgpu!news-server.csri.toronto.edu!bonnie.concordia.ca!uunet!tut.cis.ohio-state.edu!sei!fs7.ece.cmu.edu!o.gp.cs.cmu.edu!andrew.cmu.edu!<UNAUTHENTICATED>+
From: Rick.Rashid@cs.cmu.edu
Newsgroups: comp.os.mach
Subject: Re: Bytes in Mach 3.0? (mine is smaller than yours, revisited)
Message-ID: <wbke_4u00jdXMPA7Y8@cs.cmu.edu>
Date: 20 Feb 91 16:23:32 GMT
References: <62753@bbn.BBN.COM| <1991Feb15.214231.21348@watmath.waterloo.edu| <1991Feb16.002946.5711@zoo.toronto.edu|,
	<1991Feb19.175717.2996@newsserv.informatik.tu-muenchen.de>
Organization: Carnegie Mellon, Pittsburgh, PA
Lines: 61
In-Reply-To: <1991Feb19.175717.2996@newsserv.informatik.tu-muenchen.de>

Good joke.

The fact of the matter is that, in many ways, Mach provides 
substantially more functionality that a 4.3BSD or similar kernel
provides.

Specifically:

	1) VM -
	    Mach's VM is substantially more sophisticated
	    than 4.3BSD.  It provides for large, sparse
	    address spaces, copy-on-write, a well-defined
	    layering of machine independent and dependent
	    memory mapping code, etc.

	    Mach also provides for the allocation of memory
	    objects by user programs and external paging.

	2) IPC -
	    Mach provides a secure, extensible IPC facility
	    with a variety of features designed to simplify
	    or improve the implementation of multithreaded
	    servers.  Moreover, the speed of IPC is critical
	    and much complexity of code in the IPC
	    parts of Mach is due to a need for  fast
	    IPC operations.

	3) Parallelism -
	    Mach provides multiple threads per address space.

	    Mach provides support for multiprocessors and
	     multiprocessor scheduling including the dynamic
	     allocate of processors to threads.  It also supports
	     multiple scheduling classes and provides the
	     mechanism necessary for efficient implementation
 	     of multithreaded programs.

Mach (and systems like it) are best described as 
"system software kernels".  They provide key software abstractions
of hardware functionality (tasks, threads, ports, messages, memory
objects) which allow traditional system services such as file systems,
data bases, window managers, etc. to be efficiently implemented
outside the kernel.  

In fact, Mach is defined in such a way that even its own 
functionality can be implemented outside the kernel.  Memory
objects, ports, and even tasks and threads can be effectively
implemented or simulated by user programs in such a way
that no other Mach program can distinguish these "external"
objects from the internal ones.  The kernel itself is a Mach
task with multiple Mach threads of control operating within it
and it makes use many of the same IPC and VM calls 
used by user programs.

I think the key point here is not "is mine bigger or smaller than yours",
but rather, is my kernel defined and implemented in such a way
as to allow system applications to be efficiently implemented
above it.  If it is, then you can use that kernel as the basis for
many new operating systems and system applications without
having to resort to placing ever larger amounts of code in
supervisor state.