November 18, 2004

Commentary: Linux is not Red Hat, and other Sun-isms debunked

Author: Jem Matzan

Sun Microsystems head honchos Scott McNealy and Jonathan Schwartz often equate Red Hat with all of GNU/Linux. After interviewing both of them on Monday during the day-long Solaris 10 launch event in San Jose, I understood their frame of reference on this matter and many others much more clearly. For the first time in several releases, Solaris is actually a threat to the other players in the operating system market, but Sun's market outlook and publicity strategy may be working against the merits of Solaris 10.

The day began with a presentation outlining the new features of Solaris 10 and how this latest release of Sun's flagship operating system stacks up to HP/UX ("a dead operating system" according to Jonathan Schwartz), AIX ("a dying operating system," Schwartz quipped -- I was the only person in the theater who laughed), Windows 2003, and something labeled "Red Hat," as though Red Hat only offered one product to fit all situations, and all other GNU/Linux distributions were insignificant.

"Red Hat does not have military grade security, or file system innovation, or [Solaris] containers, and it's also more expensive. Red Hat requires a binary license fee per CPU," Schwartz said in summary. But Red Hat's Enterprise Linux products aren't exactly the optimal basis for comparison with Solaris 10. Other GNU/Linux distros may not have military grade security like Trusted Solaris 8, but Security Enhanced Linux (SELinux) was developed by the National Security Agency -- surely that's good enough for government work. Red Hat may not have the excellent Solaris ZFS file system, but other distributions have Reiser v4, XFS, and JFS support; it's not exactly right to suggest that GNU/Linux in general does not have any file system innovation, even if ZFS is technologically superior. GNU/Linux may not have Solaris containers (which allow applications to run in virtual instances of Solaris, isolated from the rest of the OS), but it does have Usermode Linux (UML) which provides similar functionality using a different technique. There's no argument that the now free (as in price) Solaris 10 is cheaper than Red Hat Enterprise Linux 3, but it's the same price as Fedora Core and White Box Linux, which use a more cutting-edge codebase than RHEL3 and are often used in production environments. And the bit about licensing? I don't know what Schwartz was thinking, but what he said was patently false. It would violate the GPL for Red Hat to require a binary license fee of end users. Red Hat charges for its products of course, but has neither the right nor the motivation to stop people from using it for free. The Red Hat license is the GPL, but the services connected with that software -- Red Hat Network -- are only available to paying customers. Solaris 10, on the other hand, has one of the most confusing, extensive, and restrictive licenses in the software industry. It may be free of charge, but Solaris 10 is not anywhere near free as in rights -- and that is why people say Sun is proprietary.

McNealy equates "proprietary" with "interoperable only with the same brand." While that may be true from a narrow frame of reference, the free software world tends to use a different definition; when we say "proprietary," we mean that all of the rights to that software are locked away from us. The Latin root of proprietary, after all, means "property" -- as in my property, not anyone else's. Furthermore, Sun does charge for software support for Solaris 10 -- you have to pay for an update service like Red Hat Network, and the only part of it that is not free of charge is the security updates. So in that sense it is no better than Red Hat Enterprise Linux, except it is cheaper in its initial cost. Schwartz was correct in saying that Solaris was cheaper in a certain sense, but he was less than honest in his presentation of that fact.

"I have been in this business for more than 20 years, and when a company has Wall Street, they have the market. Red Hat has Wall Street as far as Linux is concerned," McNealy told me when I asked him if he thought that SUSE Linux was a threat.

Red Hat is not the only product out there, though. There is a far superior server operating system called SUSE Linux Enterprise Server 9, and it buries Red Hat Enterprise Server 3 in every way. SUSE also has the corporate backing of industry powerhouses Novell and IBM. While Red Hat may have more market share presently, that very well may change. The irony in McNealy's argument is that he expects customers to ignore the market and go with the technologically superior Solaris 10, but he doesn't seem to think that customers would do the same with the equally impressive SUSE Linux. Sun does not have the operating system market on Wall Street -- they're not even close. So by McNealy's own reasoning, Sun is as insignificant as Novell is.

Your call is important to us

Sun's attacks on Red Hat amount to straw man tactics. Red Hat Enterprise Server 3 is nowhere near the operating system that Solaris 10 is, but RHES3 is a poor representative of what GNU/Linux can do on a server. I've already mentioned SUSE, but what about community-developed OSes? A skilled sysadmin can take a BSD variant like OpenBSD, or a community-developed GNU/Linux distribution such as Debian, Gentoo, or Slackware, and make it into a powerful server that can compete with Solaris 10. McNealy balked at that notion, saying, "And who will support that? Red Hat won't support Debian," suggesting that no company would choose to use a production operating system without corporate support from a vendor like Sun. Granted, Sun's top-tier support contracts offer some of the industry's finest corporate support. But instead of telling us about how good the company is at supporting stuff, Sun's representative hit us with more distortions and fabrications about the GNU/Linux industry. Here's the truth -- there are several companies that can and do provide corporate support for community-developed operating systems, including:

Hardware envy

Solaris 10, in the words of one of its top developers, is a collection of great, new, unique features that add up to "the world's most advanced operating system." But while it can do a lot once it's installed and running, it has nowhere near the hardware support that Red Hat Enterprise Server 3 does, let alone the amazingly diverse hardware compatibility list of SUSE Linux Enterprise Server 9. Sun just doesn't seem to care about supporting an array of different hardware. Even Sun's Java Desktop System 2, which uses an ancient version of SUSE Desktop as its base operating system, has horrible hardware support. I've struggled for months to get each successive monthly release of Solaris Express to work on my test machines; the usual foible is the integrated network chip, which is 3Com (SysKonnect/Yukon) Gigabit Ethernet LAN on one, and for a while I had trouble with an Intel Pro 1000 as well. Not only does SLES9 work famously on even the most state-of-the-art hardware from Intel, but Red Hat Desktop, which uses the same codebase as Red Hat's other Enterprise Linux solutions, works perfectly on most of my test machines where Solaris did not.

Sun does have a hardware compatibility list that contains many components and complete systems from a variety of vendors, but it pales in comparison to nearly any modern GNU/Linux distribution, commercial or otherwise.

In an interview with Schwartz, I asked him what the advantage to choosing Solaris 10 on a SPARC architecture was. "AMD and Intel aren't able to give you 16-CPU or 32-CPU systems," he said. He went on to say that some of Sun's most important hardware products were high-end enterprise-grade products that never seem to get any press attention.

The man with the golden gun

Scott McNealy mentioned the Kodak patent lawsuit that Sun settled for $92 million recently, painting Sun as a martyr that gave its money to save Java users and developers everywhere. When a journalist asked him how it saved the Java community from Kodak's wrath, McNealy tried to draw a parallel to the RIAA lawsuits against P2P file sharers. The inference that we were supposed to draw was that Kodak was going to release a salvo of lawsuits against innocent law-abiding people who had the Java virtual machine software on their computer, similar to the RIAA's notorious John Doe file-swapper lawsuits.

There is a serious flaw in this reasoning; there is no legal basis for Kodak to sue end users over their use of the JRE or JDK. End users did not infringe upon Kodak's patents -- they downloaded the Java software in good faith that it was perfectly legal, and they presumably abided by the license terms. Kodak would have absolutely no right to try to recover any damages from an end user or anyone else who was not a party to adding the allegedly infringing code to the Java source code.

"We took a $92 million bullet for the Java community," Scott McNealy told us that morning. The only bullet I see from Kodak was aimed at Sun and its billions of dollars in the bank, not innocent Java users and developers. That bullet was intended for Sun, fired at Sun, and it hit Sun in the center of the bullseye; no unintended targets were damaged. The assertion that Sun Microsystems saved anyone from the Eastman Kodak Company but themselves is best described as a highly creative interpretation of the facts.

The point of the "$92 million bullet" was to show that Sun Microsystems will "indemnify" its users against similar copyright, patent, or trade secret claims. Never mind the fact that such indemnification is pointless. Really this was a thinly veiled attack on the Linux kernel and the SCO debacle. In other words, Sun wants you to think you're safe with Solaris, and unsafe with Red Hat.

When I pressed him a bit further on the Java/Kodak issue, asking him specifically whom he was protecting from the Kodak bullet, McNealy finally gave me a good answer without any fluff, drama, or creative license: "I didn't want to confuse the Java community with pending IP litigation," he told me. That makes perfect sense -- Sun did not want developers to shy away from Java due to the Kodak lawsuit, so he settled to get it over with. I was left wondering why he didn't say that in the first place, rather than give us a dog and pony show about Kodak's legal marksmanship.

Open source: Java and Solaris

The Sun execs said some pretty silly things during the Solaris launch, but not everything was bad. Not only were the technical discussions excellent, but I managed to find out some interesting and never-before-reported details about some of Sun's licensing plans.

Someone asked Scott McNealy if Java would ever be open sourced. His answer was that he didn't think so, because he thought that the Java Community Process was a good enough solution as far as community participation and input was concerned. The problem with this answer is that it only responds to the developer end of the equation.

I asked Scott McNealy if he ever considered Java's closed licensing from a user's perspective, and I gave him the example of FreeBSD/AMD64, which has no native 64-bit JRE because Sun has not yet provided one. FreeBSD's AMD64 edition is uniquely limited because 32-bit binary support in the software is not yet fully implemented, so you can't easily use the 32-bit FreeBSD Java Runtime Environment. You can hack it to use the 32-bit Linux JRE, but that's not the point -- I didn't build a 64-bit workstation to use a hacked, emulated 32-bit Java virtual machine. Furthermore, FreeBSD users can't retrieve Java from the Ports system -- they have to visit several different sites to download binary files first.

"This is what the end user sees of Java; if it were open sourced, we could have a native 64-bit build on FreeBSD and comfortably use Java programs," I said. McNealy was silent for a moment, then said that I should talk to Software Group VP John Loiacono, who was in attendance that day but not at that particular interview. Later on I did have the chance to speak with Loiacono; he gave me details on how not-for-profit organizations like the FreeBSD project can get a free license to distribute the JRE, and suggested I read over the Java Runtime License and get back to him with an assessment of what could be changed to be more user friendly. Rest assured, I will do that. But even if FreeBSD can distribute the JRE for free, I still won't have a native 64-bit edition.

Arguably the most important question I asked Scott McNealy was, "What proprietary code had to be taken out of Solaris in preparation for open sourcing it?" McNealy responded by saying that the process of open sourcing Solaris actually started five years ago. "There were hundreds of encumbrances to open sourcing Solaris. Some of them we had to buy out, others we had to eliminate. We had to pay SCO more money so we could open the code -- I couldn't say anything about that at the time, but now I can tell you that we paid them that license fee to expand our rights to the code," he said, referring to the February 2003 multi-million-dollar purchase of expanded Unix SVR4 license rights from the SCO Group. That was at the beginning of SCO's war on Linux, and the timing of Sun's license purchase was suspicious. At the time it was widely theorized in the online press that Sun had purchased the expanded Unix licenses to help fund SCO's lawsuit against Sun's lifelong nemesis IBM and public attacks on Sun's part-time rival, GNU/Linux; if what McNealy says is true, a lot of pundits owe him an apology.

Staring at the Sun

Sun may be turning over a new leaf by open-sourcing Solaris and adopting a friendlier posture toward open source developers. Oddly enough, the company doesn't really have a reason to be so evasive; Solaris 10 is an outstanding operating system and it can stand on its own without making misleading claims about the competition.

At the end of the launch event Jonathan Schwartz made an impromptu speech; I didn't hear most of it, as I was too far away, but he did end his comments with something about Slashdotters. I ambled over to Schwartz and said, "If anyone here is going to get an article onto Slashdot, it's probably going to be me (since NewsForge and Slashdot are both part of OSTG). Tell me what you'd like Slashdot readers to know."

"Tell them that we're returning to our roots," Schwartz said, referring to the company's renewed focus on the Solaris operating environment.

"And we want developers back on our side. If there's more for us to do, we'll go do it," McNealy added. It was the first time all day that I felt that the two had broken character and simply told me what was on their minds.

Before I could thank them for their time, I was interrupted by a Sun PR flack, who informed me that I was not supposed to be there and that she was going to escort me to the door. It turns out that the press was supposed to leave a half hour before that, and that the end of the party was for Sun employees only. Somehow my colleague Chris Preimesberger and I were overlooked during the press and analyst roundup. So like the cops arresting the Monty Python cast at the end of "The Holy Grail," my colleague and I were ever so gently forced to leave the building. If only we'd been developers instead.

Originally published on The Jem Report, reprinted with permission.

Click Here!