[EMAIL PROTECTED] wrote:
The current situation - where you have massive patches like 118833-36 (a
recent 'kernel' patch) that update the kernel plus big hunks of userland
- is rather unsatisfactory.
The problem that caused 118833-36 was that multiple features affected various
subsets, and the overlaps between them resulted in combining all sorts of
normally unrelated things into the same patch. (Trusted Extensions was
really bad here, since so many different things had to be modified for
security labels.) It was recognized to be a really bad situation, and
attempts to reduce the pain were started last year to try to make things
better for S10U4 and beyond.
But that's the current Solaris patch model, in which a patch applies changes
to files in one or more packages, and is very different than the Solaris OS
upgrade model, where all packages are simaltaneously upgraded to new versions,
and very different again than the typical Linux package upgrade model, where
individual packages are upgraded to new versions, possibly requiring newer
versions of packages they depend on.
In the Linux model for instance, Trusted Extensions would have been a new
version of the kernel package that exported the new system calls, a new
version of the libc package that depended on the new kernel package and
provided the API to those new system calls, and then various command packages,
JDS & CDE packages, all with dependencies on newer libc package to get the
new functions needed for labeled systems.
--
-Alan Coopersmith- [EMAIL PROTECTED]
Sun Microsystems, Inc. - X Window System Engineering
_______________________________________________
indiana-discuss mailing list
[email protected]
http://opensolaris.org/mailman/listinfo/indiana-discuss
