Firstly, I have administered various Solaris boxes. Mostly an assortment of SPARC and UltraSPARC workstations, but also a dual USPARC E450 (very nice machine, if you can afford it). My experience is with Solaris 6, 7, and 8.

Solaris is an excellent proprietary UNIX. It is very stable, well maintained, and can be made very secure. You can bet your enterprise on it.

My negative comments are as follows. The config file design in /etc is crufty at best. The rc script design is atrocious, particularly when compared with the fine granularity of the rc script design of Red Hat.

Like all UNIX variants and clones, one is faced with the "it's all the same, only different" conundrum. All of the proprietary UNIXes have diverged, making the emphasis on the "only different". If you have administered many different Linux distros, the differences among them are minor compared to the differences between proprietary UNIX variants.

Solaris has the same problem that most proprietary UNIX variants all have, they come from companies that have marketing departments that decide what the feature list shall be (this is not always a bad thing, but it certainly can be when the company decision makers put more weight on the marketing department's input than on the input from engineering). In all fairness to SUN, I have observed over the years that they do seem to reach a well balanced consensus between marketing and engineering. This is why they have been so successful for the past two decades. However, the Linux development process places far more weight on the technical input than happens at any company, no matter how well run that company is.

I have seen no comparative benchmarks for Solaris on x86 versus Linux. Neither have I seen any comparative benchmarks for Solaris on iAMD64 versus Linux. SUN announced that Solaris 10 is faster than Linux. The fine print is that this 'faster' is on a 128 way ultraSPARC SMP box and it was compared against Linux 2.4.21. I wonder how it compares against 2.6.x?

Solaris has no advantage whatsoever on single processor machines.

Now for the reality injection. Earth to SUN, "OSes are now a commodity, we don't need Solaris anymore."

In my opinion, SUN needs to do the following in order to remain relevant:

• Continue to make excellent hardware as you have always done. The SPARC processor is not dead yet. Your new multi-core SPARC designs look to be kick-ass. You have an excellent history in building very reliable and high performance SMP big iron. Covering your bets with Opteron is a very smart move. Please keep this stuff up.
• Your support has always been excellent. Make it even better. Carly is destroying this at HPQ (I no longer have any hope for them at all), so we know that your competition in the big iron market is really only IBM and they are doing everything they can to make their support the best in the world. Follow suit. Competition is a good thing.
• It's time to end of life the proprietary OS crap. Put the effort into making Linux run really well on all of your iron, both big and little. Get all of your people to be Linux experts now. The clock is ticking, and I'm afraid you may already be too late.
• Move all of your enterprise features into Linux. Put your middleware on Linux. Stop trying to convince the world that they really want Solaris instead. It's not working. Your customers are not idiots. The OS is no longer a factor in competing. You have to compete on middleware, enterprise level features, and support. Using Linux everywhere will actually make it easier to steal customers from your competition.
• That Linux distro that you market as the "SUN Java Desktop", how about putting the L word in there somewhere.
• Did I mention "GTF over your Linux phobia"? Is it time for Scott to retire? Who is making these bogus decisions?

I like competition. Competition means I have choices. SGI is now selling 512 Itanium Altix machines running Linux and so they may yet survive. IBM is putting Linux on every machine they sell. Both of those companies have end of lifed their proprietary UNIXes, but will of course support any of their customers that choose to remain on those platforms for quite some time. HPQ is now becoming a dead duck in the enterprise datacenter (IMHO, it should be a criminal offense to have management that stupid).

It will be a very sad day if SUN goes the way of Digital Equipment Corp, but I fear it may well be that I am watching history repeat itself. It's rather frightening how much damage arrogance can cause.

"Cause if it ain't broke, I don't fix it."

"But Solaris has predictive self-healing."

"Our Linux server hasn't ever needed healing."

"Solaris has Extreme Network Performance!"

"Great! Will it speed up the printer?"

"And Solaris enables you to give each application its own private environment, so you can reboot a container for whatever reason."

"We've never rebooted the Linux server."

"Solaris can run Linux applications natively."

"Well, the fine print says we have to have Linux on the server anyway."

"Feh, I think you're just being negative. See this glass? Do you see it as half-empty or half-full?"

I just have to ask all of the Linux geeks - will you use it?

Yes, of course.

So when your boss says "Why aren't we using Solaris on our x86 boxes instead of Linux", what are you going to say?

Schwarz himself has criticized the U.S. patent system, but at the same time, he has applied for a patent on Sun's per-person, per-year pricing.

At home, no. The hardware is mostly beige-box x86, and I've used linux for long enough that when I build (or re-build or re-install OS on) a new computer, linux is the default choice, with OpenBSD on the firewall.

At work, sure. Our customers aren't computer geeks, so when they buy a computer, they won't mess with the installed OS; on a Sun machine, that'll be solaris. Our software has to work there (and on other OSs). But (a) I don't admin those computers; and (b) as far as our software is concerned, there isn't too much that's specific to solaris or linux or whatever (and what is there, we try hard to keep segregated in a few portability areas).

I don't think open-sourcing solaris will slow linux; I think there are enough more people working on linux (that is, specifically kernel developers) that it will evolve faster than solaris will, and as someone else said, linux development is much more tilted toward the technical side, which is something we linux geeks like.

No.