Re: The Kernel (fwd)

["Star's End" <starsend@interlog.com>]

> 
> George Bonser wrote:
> > 
> > No, I was not.  Once you add a new kernel, you are no longer SEUL. There
> > is no longer a guarantee that everything compiled agains the original
> > kernel will work.  It is one of the reasons that people are so afraid to
> > produce anything for Linux.  The user keeps dicking with the OS.
> 
> That doesn't make a bit of sense.  I can understand not wanting any
> pathes to the main part of the kernel for the core.  That would alter
> the functionality of things and then, potentially, cause problems for
> vendors.  That is why there needs to be a stripped kernel in core, so
> the bare essentials are standardized across the board.

This really is a mute point. The people targeted by SEUL will probably
never recompile the kernel. And if they do it is there problem.

My humble suggestion is a minimal kernel with everything compiled as 
modules, using kerneld.

If this causes problems then you solve them individually. Hopefully the 
problems will be rare.

> 
> Core is *not* SEUL.  At least acording to what has been discussed
> concerning a "standard core" is not SEUL.  SEUL builds on top of this
> core just like anybody else.  That means different kernels can be placed
> in *base* (the layer above core) because that is where the dist begins.
> 
> You aren't going to tell me that if a kernel with, ipmasq compiled in,
> is in base that it will screw up vendors and therefore can't be there. 
> Kernel binaries with different options compiled in can and should be in
> base.

To the best of my knowledge the only problems an application has at the
kernel level is the a.out vs elf format or if a module is missing. Anything
else takes up memory, but doesn't affect an application.

> 
> > I do not see IBM releasing their source, or Microsoft, or anyone else yet
> > there are plenty of apps. Why?  Because they have a clear API for the
> > programmer to work with.  You start recompiling the kernel and your API
> > starts moving around on you.

This is a mute point. Most users of SEUL are never going to recompile the
kernel. 


> 
> There are actually 2 projects going on here.  The Linux core proposal
> and the SEUL distribution.  They have to remain seperate concepts.
> 
> Patches to the kernel are a different matter.  But a beneficial patch
> should be concidered for inclusion into the Linux core as well as SEUL
> base.
> 
> 
This is where you lost me. What is a beneficial patch to the kernel?
I think we are missing the point here. The following is the concepts I
have gotten out.

	The linux kernel - the operating system - scheduler, device drivers
				etc ...

	The linux core - the minimum required. This includes a filesystem,
			and the programs in /bin /usr/bin, the libraries in
			/lib, and /usr/lib, and other things like the X11 
			subdirectory ...

	SEUL - the extra applications that a user may want but are not 
		required by the linux core.


Know people have expressed fears that if a kernel compile is performed
by the user the installed applications may break. This is always a danger
when compiling the kernel. there is no way you can solve this problem.

The correct problem to solve is the following. Given a set of applications,
what should be included in the kernel such that they work out of the box?

B.T.