On Tue, Dec 11, 2012 at 9:03 AM, Irek Szczesniak <[email protected]>wrote:

> On Fri, Nov 30, 2012 at 1:13 PM, Joerg Schilling
> <[email protected]> wrote:
> > Sriram Narayanan <[email protected]> wrote:
> >
> >> I should clarify, I decided to address release engineering first, since
> the
> >> whole non-oracle Solaris community will need to get together to solve
> the
> >> problem of tracking package changes, have the tools to solve posix
> >> compliance debates,  and then start to write device drivers.
> >
> > Well, this is something I mention since years. I still hope there is a
> way to
> > collaborate - note that even the *BSDs that have been disintegrated
> since a
> > long time do important things together.
> >
> > What we of course first need is to decide on a minimal set of consensus.
>
> Which minimum set of consensus? The point is to go boldly forward with the
> userland modernization where Illumos has failed with it's conservative
> policy.


If there was failure, it was with OpenSolaris programmers being too
aggressive, and the rest of us inheriting the mess.

Core OS functionality (booting, services, packaging, etc.) should not be
dependent upon newer GNU userland, but based upon thoroughly debugged
Solaris userland. These binaries should be separate and distinct
locations (perhaps: /bin, /sbin, etc.) for simple maintenance and to enable
easy embedding into smaller form-factors.

If people wish to have a newer userland, where bugs are constantly being
thrashed out, from other developers in other spheres (not interested in
Solaris longevity) - then those should be put elsewhere and people should
have the option to pick & choose (perhaps: /usr/bin, /usr/sbin,
/usr/gnu/bin, /usr/fbsd, etc.) and those communities inherit the bugs those
userlands suffer from. The core should be wary of suffering from external
bugs that others may not approve fixes for or suffering from feature
suppression where others may block innovation regarding.

There is soooo much that is plain-old broken in gawk, even though it has
some GREAT additional features, virtually nothing coded in any of the
standards related awk will run in gawk. A standards-compliant userland is
important for software compatibility. A modern userland is important for
end-user comforth. Both are aggressively needed.

Sun and Oracle, in OpenSolaris and Solaris 11, merged the standards-base
and gnu-base userlands together without strict differentiation,
contaminating the core OS services, which was perhaps the poorest
engineering decision I have ever seen. I would hope others do not make the
same mistake.

The purpose of an OS is to run software - if existing commercial software
will not run under the OpenSolaris splinter, there may be little reason for
the splinter to exist, with the exception of some special purpose
appliance. The usefulness of a special-purpose appliance running with an
OpenSolaris kernel without USB3, WiFi, or clustered-ZFS support is puzzling
to me... unless it is embedded - and then modernized userland becomes less
important.

Clearly, Illumos is driving OpenSolaris source code towards storage
appliance and cloud-based hypervisor. Few in those groups understand the
necessity of clustered ZFS for storage to provide good any-to-any H-A or
D-R... this tells me those projects may not be long-for-this-world.

I guess the rest of us need to decide what WE WANT out of an OpenSolaris
splinter.

I want a basic Desktop OS running SunRay Thin Clients, run some commercial
apps, against a shared-nothing clustered back-end hypervisor. That's it.
USB3 is a bonus, for faster external storage. WiFi would be a bonus for
wireless storage. If no sunrays, then I need a cheap out-of-the-box
thin-client alternative. I want support for SVR4 standards and a community
effort to drive innovation (superset) in this arena.

Thanks - Dave
http://svr4.blogspot.com/
_______________________________________________
belenix-discuss mailing list
http://mail.opensolaris.org/mailman/listinfo/belenix-discuss
http://groups.google.com/group/belenix-discuss

Reply via email to