Last week, we took a stab at prepping readers for this week's LinuxWorld Expo in New York City. We failed, however, to make the two guaranteed slam-dunk predictions we could make every year: Somewhere, someone is going to say "this is the year Linux finally arrived in the enterprise" and that someone, somewhere is going to say "this is the year Linux will finally arrive on the desktop."
So much for an easy opportunity to look prescient and nail down a consulting gig somewhere. All the same, a lot of news did come out of LinuxWorld that helps make the case, even if people have been making it since Oracle turned up on Linux several years ago.
We believe, however, that the cautious voices from within the Linux community itself are more indicative about how mature Linux has become than the array of product releases, partnerships, and certifications we will highlight later in this column.
A prime example of this theory is the hesitation surrounding the latest kernel release.
Several weeks ago, we considered a minor tempest over the new Linux 2.6 kernel. The long and short of the issue was concern over the effect a new kernel release, and the subsequent freeze on development of the 2.4 kernel, would have on the overall project. We didn't see a lot to worry about because when it comes to refining Linux, we believe much of the work is going on in places outside of the very public mailing list arena of the project proper.
At the same time, we raised, and subsequently walked past, a more troubling issue that came up the last time a new Linux kernel was released: the disconnect between the needs of Linux enthusiasts (the gratification of playing with something new) and enterprises (stability).
When Linux 2.4 was released, an admin of our acquaintance was so overcome with the "thrill of the new" that he executed an upgrade from the old 2.2 kernel to the new 2.4 version before a single point release came out -- on production servers. An informal poll of other Linux users left us unsettled: Several had done the same; others just hadn't gotten around to it but would soon.
At the risk of having to turn in our Linux Fan Club decoder ring, we'll admit that even now, a few years down the road, this strikes us as one of the sillier things we've heard. It's almost as silly as the part-time admin who told us that using Linux as a Web server was impossible since with the advent of the graphical browser Unix variants would start crashing because they were text-based.
So we were relieved to hear a voice from within the Linux community advocating caution about the new kernel and the implications of upgrading right away.
The Open Source Development Lab, an Oregon-based consortium, is very involved with prepping Linux for enterprise deployment. Its director, Tim Witham, was on hand at LinuxWorld. He noted that the thing to be doing with a new Linux kernel at the moment is testing, not deploying.
Throughout the years we've heard people grumble that the continual development Linux undergoes out in the open is discomfiting and nervewracking. We beg to differ. With some software companies, which we will refrain from naming, an opaque development process combined with a sense that the profferred upgrades and patches tend to be all-or-nothing affairs is a lot more troubling. During the next several months, and throughout the year, a moderately dedicated admin or IT manager will be able to watch Linux vendors and notes for Linux point releases (which our sister site, LinuxToday, posts with regularity) and make informed choices on when this kernel is ripe for deployment.
In Other News
We've tussled with the question of IP indemnification more than once, and come down somewhere in the middle: It's a nice guarantee if you can get it. IBM, which is embroiled in a suit with SCO over intellectual property issues, remains steadfast in its insistence that indemnification isn't necessary.
This week, the company stuck to its guns. Red Hat, which has also avoided offering indemnification, did, too, but in a less pointed manner. The company announced the Open Source Assurance Program (OSAP). The key feature of OSAP is what Red Hat calls an Intellectual Property Warranty. It doesn't offer indemnification but rather a promise that if infringing code is found in something Red Hat sells, Red Hat will sanitize its product of the infringement.
We see one possible snag in its logic. Part of the whole issue with SCO's numerous suits and general conduct has been a disinterest in identifying infringing code and allowing the alleged infringer a chance to clean it up. It's an easy distinction for Red Hat to miss since, unlike SCO, Red Hat makes money from selling things, not litigating over them.
Seen and Heard at LinuxWorld Expo
This week's parade of announcements about software and hardware from LinuxWorld has the pundits talking about how "Linux is now enterprise ready" (a statement as predictable as kabuki and has been since, well, the last century). Here are some highlights to come out of the show:
- IBM is preparing to push Linux on the PowerPC platform in a move to expand its BladeCenter offerings.
- Big Blue is also making a case for transitioning from older Windows NT 4 deployments to Linux.
- Novell made a a raft of announcements:
- The beta version of GroupWise 6.5 for Linux is out, tying together recent acquiree Ximian's Evolution PIM with the messaging system.
- Ximian's Red Carpet, which started life as a software management tool for desktop users and has been in the process of evolving ever since, now works on IBM eSeries servers.
- Novell has left the UnitedLinux consortium, and taken recent acquiree SUSE with it, in favor of working with the IBM-led Eclipse Project, which provides a cross-platform IDE. UnitedLinux, which started life as a desperate holding action against Red Hat's dominance, is down to SCO plus regional players Conectiva (in South America) and TurboLinux (in Japan and the Pacific region).
- SUSE Enterprise Server 8 has achieved ELA3+ certification, which makes the distribution more competitive for certain government contracts, and puts it in the running with offerings from Sun and Microsoft. IBM sponsored the certification.
- Xandros announced the release of xDMS, an "enterprise desktop management server." Once upon a time, the core Xandros distribution was a Corel product built on top of the Debian distribution. xDMS represents a good direction for a "desktop Linux" company these days: It stresses the enterprise value proposition without making a lot of wooly-headed claims about a Linux desktop under every Christmas tree this coming year.
- Penguin Computing, long in the Linux hardware business, announced "the first 64-bit implementation of the Linux-based Scyld Beowulf cluster operating system." The OS will be available for both AMD Opteron and Intel Itanium 2 systems. Opteron continues to make inroads in Linux-land, as the company also announced five new Opteron-based servers for clusters and one new Opteron-based workstation.
- Veritas announced the availability of Veritas Cluster Server and Veritas Foundation Suite for SUSE Linux Enterprise Server.
In SCO news, several sites around the Web linked to a leaked letter, (that's a 370 KB PDF) apparently meant for U.S. legislators. Hopefully the likely download time or the clumsiness of reading PDFs in a browser window will scare most people off. We've gotta report this stuff, but as we've stressed time and again, SCO's public denunciations of Linux and the GNU Public License veer between distortion and unintentional McCarthy-era kitsch. Little in this letter is different from previous assertions.
SCO also launched a "slander of title" lawsuit against Novell for its claim that SCO doesn't own as much of UNIX as it claims it does.
Everyone busy at LinuxWorld made for a quiet week on the security front. We did, however, catch one widespread patch for slocate, a secure replacement for the locate command, which allows users to find files by searching a regularly updated index instead of crawling through the file system each time with a find command. Several vendors have reported patches. The bug could allow a malicious local user to gain access to every entry in the slocate database, providing potentially compromising information.
Tips of the Trade
Highlighting the news coming out of LinuxWorld has left us short of space, so a quick tip this week:
You may remember our tip about how to use find to locate large files. Well there's another handy tool that does the same thing: du, which is more single-minded than find and, perhaps not incidentally, a little easier to master.
du returns the file size of every file beneath the current working directory and takes quite a few arguments to modify that basic behavior. The most useful are the -h switch to specify "human-readable" size reporting (KBs and MBs instead of blocks), the -c switch to report the grand total of all files in a given hierarchy, and the --max-depth argument to limit how far down the directory tree du searches.
This week, Enterprise Unix Roundup offers a tip of the hat to regular tip contributor Ed Heil.