On Wed, Oct 22, 2003 at 10:47:17AM -0700, Cory Petkovsek wrote:
> On Wed, Oct 22, 2003 at 10:13:52AM -0701, Jacob Meuser wrote:
> > My comments:
> > I wouldn't be surprised if he had problems with OpenBSD-current, since
> > he used a -current from shortly after the tree unfroze for the next
> > release, which is when most "new" things go in and OpenBSD-current is
> > most unstable.
> So these "new" additions are to the kernel and are so drastic in an
> incremental release as to introduce stability problems and O(n)
> algorithms as opposed to O(1) algorithms used in free bsd 5.1 and linux
> 2.6?

Definitely stability issues.  Some of the work being done is on
threading and random address/page mapping etc.  Does his Leanux kernel
have PaX?

OpenBSD really has only one source branch, not separate stable and
devel branches, releases are meerly snapshots of the code, and -stable
just has patches backported from -current.  So, yes, new stuff goes in
right after a release, so that if the major bugs can be fixed within a
few weeks, it can stay in for 5 months of testing and fixing before the
next release, otherwise it is removed.

How else would one incorporate major changes?

> > And he surely has a more "standard" machine for server testing than a
> > laptop, no?  
> Indeed.  However it is the same hardware for all platforms.

But only OpenBSD was installed in what was a swap partition for another
OS.

> > (OpenBSD will try to configure any device it finds, as
> > almost all drivers are built in to the GENERIC kernel ... I doubt he
> > loaded all possible modules on the other OSes.  Not saying this is
> > necessarily a problem, but he doesn't even give the dmesg for each
> > OS, or if IRQs were shared with some kernels and not others, etc, etc.)
> Nevertheless, what's wrong with his suggestion of having openbsd
> incorporate many of the kernel improvements found in netbsd and freebsd?

None, but he doesn't need possibly unscientific data to make such
a suggestion, which has been made many times before.  UBC was actually
incorporated in OpenBSD for a brief period of time (right after
3.2 release I believe), but it caused too many problems and was
removed.

> And his response to the response "openbsd's focus is security not
> performance."  "Hopefully security and performance are not mutually
> exclusive."  With Openbsd's security track record and freebsd's
> performance record, what a gain it would be to incorporate freebsd's
> improvements into openbsd.

Maybe.  Or maybe FreeBSD could incorporate OpenBSD's security
improvements.  There's probably a reason neither has happened yet.

> It is easier to make any OS higher
> performance than it is to make any more secure.

Well, FreeBSD and OpenBSD are quite different.  There would be a fair
amount of porting involved, and then the ported parts would need to
be analyzed so they don't create new security issues.

Also, it's not a matter of security and performance being mutually
exclusive, but that no OpenBSD developer has decided to make performance
his main goal, and who is to say what a volunteer is to spend his
time on?

-- 
<[EMAIL PROTECTED]>
_______________________________________________
EuG-LUG mailing list
[EMAIL PROTECTED]
http://mailman.efn.org/cgi-bin/listinfo/eug-lug

Reply via email to