Enterprise Unix Roundup: A Red Hat Retrospective
You've probably read a lot about Red Hat's latest Linux release already, so we're not going to go down the laundry list of features for you here. It's enough to say that with support for 32-way servers, address space limitations that have soared to 256GB, and support for big iron across several platforms, the company's taken another step forward in what seems to be an endless Linux process: arriving. Red Hat is earning praise this week for its latest Enterprise Linux release, which makes it a perfect time to look back at a company that gave away product while it outwaited the competition. And don't fear the Panther: OS X's move to postfix brings a sane sendmail alternative.
What's the most interesting about Red Hat's news this week is how with the release of its Enterprise Linux 3 product and recent changes in approach Red Hat is looking like a 21st century Linux company more than it ever has.
We remember our first Red Hat experience with some fondness. It involved Red Hat 4.2 (released in May 1997), and after some time spent slaving away in older, creakier distributions, much about the way Red Hat went about its business seemed like a revelation. It was clearly a company bent on taming Linux. It's a comment on the state of the Linux world at that point that Red Hat was regularly decried as having some sort of corrupting influence thanks to the way it made installing software, for instance, easier than seemed decent to Unix graybeards and assorted other neo-Calvinists.
The fact was, Red Hat made stuff “just work” that should just work: Getting X Window up and running, for instance, wasn't a heinous test of will and perseverance. Getting a database installed became the easy part of a process that is rightfully back-loaded with hassles during configuration. All automation, which Red Hat provided in spades, is fraught with some peril, and there's no denying that the company earned its reputation for putting out “.0” releases best avoided by anyone who needed to do more than wait around for the inevitable raft of patches.
Despite its occasional stumble, though, Red Hat became the putative parent of much of the Linux boom in 1998. Its biggest competition among hobbyists was a near-carbon-copy derivative (Mandrake), and its enterprise competitors included companies happy to borrow its packaging tools (SuSE) and most of the rest of it (Caldera). During that period, people began to ask the logical questions about Linux companies, which were springing up like mushrooms: What sort of money could they make selling something people could get for free? What kind of money could they possibly make selling something they gave away for free themselves? The patient answer was always “we'll sell service and support,” which seemed to be enough of an answer to catapult Red Hat (and a few others) to insanely inflated IPOs.
With the dotcom bust, that answer lost a lot of credence. While people agreed Linux was robust and useful, nothing about it where an enterprise audience was concerned seemed sophisticated enough to justify benching the in-house Unix talent. All the while, the company was giving away product and trying to make a living off of $39.95 box sets and phone support deals.
It's blindingly obvious now that as business models go, that one was terrible. One could argue, in fact, that the only sense it made was as a way of loss-leading with three years worth of product for the sake of establishing brand. We'll admit to sipping a little of the kool-aid along with the rest of the “service-and-support” mantra-chanters, but Ronald Reagan used to vote Democrat, too.
By 1999 and 2000, the company had moved firmly into its “Red Hat Network” stage, a period during which not a lot of value was ascribed to its distribution, thanks to an emphasis on its maintenance software, which automated mass upgrades and system management for large organizations. Perhaps its most valuable commodity during that time, though, wasn't decent product or interesting revenue ideas: it was patience.
Over the past several years, the company has maintained its brand and done the things a company has to do to look like money: It has made itself “The Linux,” pushing Caldera aside and holding the line against SuSE in all but highly specialized markets in the United States even as those companies partnered against it in an attempt to loosen its hold on North America. But as much as the average enthusiast might feel compelled to take a swipe at the undisputed leader in enterprise Linux, we've met precious few people in enterprise IT who will consider any other distribution: it's the one with the software partners, the name, and the reputation.
We're confident that when the history of Linux is written, Red Hat will get a lot of the credit for keeping a foot wedged in the enterprise door for other Linux distributors to follow.
This week, with the release of its latest Enterprise Linux distribution, Red Hat moves firmly into a stage becoming of a high-end software company: Its reliance on retail boxes has been forsworn, its focus is on its enterprise offerings, and it's in plainly in the business of selling an operating system you can trust your business to.
Yes, Linux is “free,” providing the raw materials for an operating system worthy of enterprise computing. Theoretically anyone can do what Red Hat is doing with those raw materials, but not many companies have. If you have any doubts about the company's place, consider the grudging respect it gets from Sun (reported here last week), where it is clearly causing some heartburn as that company tries to revive its own x86 offering from an earned reputation for anemia.
Red Hat made it easy just to get a mouse to work six years ago is scaring Sun out of its wits and establishing Linux as an enterprise powerhouse today. Not bad work.
In Other News:
Speaking of doomed business models and questionable giveaways, after two years of searching for free penguins at Linux trade shows, we finally managed to score not one, but two, at Jupitermedia's Enterprise Linux Forum in Washington, D.C. this week. We're not sure if it's evidence that the economy is on the fast rebound, or confirmation that Linux has truly ascended to the enterprise class. The recurring theme of the show was Linux's changing place in the enterprise. Examples of companies with widespread Linux deployments were much cited (amazon.com and Merrill Lynch being the two most prominent), and several sessions focused on scaling up or clustering. The big elephant in the conference center was the SCO question. When asked about it during the question session following his keynote, Red Hat CTO Michael Tiemann's response was that Red Hat had "filed opinions in a Delaware court." When pressed further, he suggested the attendee check out Groklaw, which has made a cottage industry out of keeping up with SCO.
Speaking of SCO, the company has announced that it isn't interested in the small fry with its Linux licensing scheme: It's content to stick to the Fortune 1000 for now, and says the previous non-panic it sent through companies with its threats to start sending invoices was a simple miscommunication.
Sun and Fujitsu, which also uses the SPARC architecture, are reportedly in talks to unify their server designs and merge production of their high-end offerings by 2004. The Japanese paper Nihon Keizai Shimbun reports the companies have reached a tentative agreement, and that their combined might could spell control of more than 40 percent of the Unix server market. Neither company had much in the way of comment, except to admit that they're talking.
Another Unix hits a milestone this week as Apple's OS X 10.3 (“Panther”) is released.
It was a quiet week for security problems, but we caught one from Red Hat that put us in the mood to pontificate: A recent kernel update broke the iptables utility, and, by extension, iptables firewalls under Red Hat 7.1-8.0 and Sun Linux 5.0.7. The moral of the story? This is why automatic update utilities like Red Hat's own up2date and Ximian's Red Carpet don't go too near the kernel when they're working their magic and making it easier to slip out of the office before 5. Too much can go wrong.
We've been there, and we know the temptation to just give a kernel update a pass is strong, but it's a temptation worth resisting, as this problem illustrates. Your vendor's human, too. Or staffed by them, anyhow, and they make mistakes that can put you in a world of hurt if you rely on their infallibility.
Tips of the Trade
One of OS X 10.3's new features is the introduction of postfix, which will replace sendmail. A few of our colleagues who depend on OS X servers in their organizations have expressed a little vexation with the change. As our review on it noted, though, there's not a lot for a competent admin to worry about. We recently migrated to postfix, and found hardening it against spammers and doing a lot of basic maintenance to be a real pleasure compared to sendmail and even the occasionally baroque eximWhere sendmail has provoked fear, trembling, and the use of a macro language just to configure it, postfix is a relatively amiable piece of software for a mail transport agent (MTA).
Our best advice, if you're looking at making the move to postfix: check out the documentation on the project site, and remember that if you make a configuration change, the command postfix reload will spare you needing to restart the entire service.