On 6/17/2010 7:33 AM, Brian Utterback wrote:
On 06/17/10 02:35, Milan Jurik wrote:
The reason is simple for improving "desktop feeling" and you heard it
here already - people prefer to use system which they know, on both
sides. And not every admin will maintain own desktop at home for himself
and the other for the rest of his family. I am not saying we should push
commercial games for OpenSolaris. But we should try to make OpenSolaris
good enough for typical users also. Otherwise only small portion of
people will know about Solaris and will use Solaris on universities and
will use it in their datacenters.
If OpenSolaris will end in niche market only then it is on way to die.
All niche markets are shrinking over time.
There is need for an "OpenSolaris-mostly-everywhere" strategy in
long-term. Otherwise it will be "OpenSolaris-nowhere" sooner or later.
This was the argument for the Solaris Modernization/Familiarity
project and for giving it away for free. But that didn't work. Of
course, we have two schools of thought now, one is that Sun failed
because it abandoned its "roots", and the other because the change was
"too little, too late".
The reality is that using the same OS on your desktop as on your
server is a benefit like any other, to be weighed with the whole
package of costs and benefits. If there is a compelling benefit for
them to be different then the industry will adopt different OS's.
I think that Oracle management thinks that they can leverage the
Solaris enterprise features and closer integration with other Oracle
products into a successful enterprise stack. Expect to hear the mantra
"Runs best on Solaris and Oracle servers" soon. If even a significant
portion of current Oracle customers are persuaded to adopt the whole
Oracle stack, then the the purchase of Sun will have been a success.
So, does Oracle need to make a better desktop to accomplish that? I
don't think anyone really knows and they haven't really decided yet.
The same question applies to the whole Open Solaris project, I think.
I am not privy to any inside information about the thinking of Oracle
management (I wish!), these are just my own thoughts on the subject.
The problem here is limited resources. Even communities like Linux, or
huge software shops like Microsoft has the problem, let alone smaller
ones like the various *BSDs and OpenSolaris.
Where do we put our limited efforts? If we put them in A, we can't do
both A & B. It's that simple.
Take Linux for example. Right now, the kernel supports a vast array of
hardware. There are lots of very nice desktop-oriented features in the
kernel (good USB & firewire hot-plug support, very light-weight process
forking, relatively straightforward kernel tuning) plus plenty of
userland support for this. However, I can give you a laundry list of
other features that aren't small machine-oriented that are at best
poorly implemented, or missing altogether, and that no-one really is
working on fixing. The Virtual Memory subsystem seems to be a member
of the re-write-of-the-week club. There *are no* reliable modern
filesystems. Volume Management is a horrible mess and bug-prone.
Scaleability is slowly getting better, but still considerably behind
state-of-the-art.
None of this is to disparage Linux as a whole - it's just the current
state of affairs. Running on commodity x86 hardware is the big focus.
Features which aren't available on your $500 desktop machine have
radically less developer support throughout the community. The sheer
number of Linux developers hides this, but take a look at the number of
people working on the VM, processor scaling, and system management
features, then take a look at the number working on consumer device
drivers, media players, and desktop window managers.
Also, ask yourself this question: the *BSDs have pretty much just as
much "friendly desktop" support as Linux. Other than a more limited
hardware driver support, they can do pretty much everything a Linux
distro can. They're freely available. They have excellent OS services
for desktop usage. And a developer community considerably larger than
OpenSolaris, with the ultimate in "business-friendly" licensing.
Yet, they have virtually no desktop usage penetration. Extremely limited
ISV support.
So, if the *BSDs can't do it with several times OpenSolaris' community
behind them, how are we supposed to succeed? Sheer willpower?
Everyone also seems to forget this isn't a zero-sum game. Linux isn't a
monolithic OS - the differences between the competing distros is about
as large as the differences between a Linux distro and OpenSolaris. Try
throwing someone whose only ever administered a RHEL server and give
them a Debian box. Watch them flounder. For days. Most of the commands
are the same, but many of the critical ones aren't. Locations are all
different. Packaging is radically different. Installation is radically
different. Systems management is totally different. It's no harder to
get that person to learn OpenSolaris than a completely different Linux
distro. And, therein lies our salvation, as well as our competitor.
Like my previous post said, we don't need to have OpenSolaris on kid's
desktops. Linux gives us the "in" by familiarizing the population with
the "UNIX" mindset and paradigm. OpenSolaris then gets to leverage this
knowledge in a new direction, if we can get to them at the time when
they've learned enough to be able to see the potential advantages of
OpenSolaris; in my opinion, this is at University. Getting OpenSolaris
to Universities is critical; there it is most likely to have the
appropriate hardware (not necessarily Oracle-brand) able to showcase
OpenSolaris' capabilities. If you can't showcase the differences in a
meaningful way, why would people chose OpenSolaris over Linux, anyway,
regardless of earlier "exposure".
Sure, many groups pick the OS they are most familiar with. This is
particularly true for small businesses, which have very limited
resources to spend in IT. On the other hand, "black-box" appliances are
extremely appealing to this market segment, and OS choice is irrelevant
to their success. And for Big Business, familiarity is but a single
bullet point on the list of requirements. Remember, we're talking about
companies that will do forklift upgrades from AIX/Power to SPARC/Solaris
(or vice versa) or similar infrastructure changes if the benefits are
there.
We're arguing the same point that Commoditized (i.e Whitebox) vs Custom
(i.e. brand-name) hardware people make all the time. It turns out
neither is right - while we find more and more usages for true commodity
hardware, we also find that custom hardware retains huge advantages that
make it much more appealing in many settings. Sure, I don't see people
buying a new IBM x-series machine for home server usage. But, how many
companies run their critical infrastructure on non-brand-name
equipment? Oh, right, no one except the massive parallel data services
(Google, Yahoo, etc.). Both have their place, and neither is suffering
for it now.
--
Erik Trimble
Java System Support
Mailstop: usca22-123
Phone: x17195
Santa Clara, CA
_______________________________________________
opensolaris-discuss mailing list
opensolaris-discuss@opensolaris.org