Multiple logical CPUs on a single physical CPU, or hyperthreading (HT), is a relatively new concept introduced with Intel's Pentium 4 Xeon that lately has been getting a huge amount of support in the Linux kernel. Ingo Molnar started it off in late August 2002, with a patch to add HT-awareness to the scheduler. Many patches have followed since then, and hyperthreading has become quite the mainstream feature.
khttpd is finally going away. After a long and bitter struggle from the moment of its inclusion, the kernel-based web server is gone from 2.5 and will not be back. In spite of the tremendous controversy surrounding this feature, it was not the flame wars that eventually sealed its doom, but the fact that user-space Tux2 is much faster. There is even some talk of putting Tux2 into the kernel as a replacement for khttpd, though many folks object to this on the same grounds that they originally objected to khttpd. Some patent issues also are holding up such an idea, and it looks as though no one really wants to bother fighting it out.
Several new system calls found their way into the kernel in August and September 2002. Among them, clone_startup() on x86 boxes reduces the number of system calls required for thread creation to one. Glibc's fancy pthread code is one big user of this call. The only problem is that the name has not quite been nailed down yet. YMMV.
One cute little tidbit: the PC speaker in post-2.5.31 kernels may now be used as a microphone. This is new and weird. As Jos Hulzink put it on the linux-kernel mailing list, “2.5.32 will go into the history books as the kernel that implemented voice recognition for all AT class computers....”
The struggle for a new kernel configuration system is ongoing. With CML2 apparently out of the running, several new configuration systems have emerged. Among them, “kernel conf” may be the most likely to succeed. Roman Zippel has been working on it steadily and claims his code is nearly completely usable. The rumor is kernel conf is likely to go into the main 2.5 tree, but no official announcement has come out yet.
A new tool, called devlabel, has surfaced from Gary Lerhaupt. It allows consistent access to storage devices via dynamic symlinks, with support for hot plugging. Simply plug in your device, and a symlink appears that may be used to access the device. Unplug the device and the symlink goes away.
Number of wireless nodes discovered with an iPAQ by hackers flying in a four-seater Grumman over Perth, Australia: 95
Number of wireless nodes discovered by the same hackers using a Toshiba laptop: 92
Speed in MPH of the Grumman over the ground: 250
Altitude in feet of the Grumman throughout the flight: 1,500
Number of wireless nodes discovered by Phil Windley, CIO of Utah, flying in a Piper Turbo Arrow over Salt Lake City: 27
Number of encrypted nodes among those: 5
Speed in MPH of the Piper over the ground: 125
Altitude in feet of the Piper throughout the flight: 1,500
Linux percentage share of 110-million-unit desktop market: 2.7
Unit sales growth of Linux desktops in 2001: 47
Numbers of Linux desktops distributed for every one purchased: 12-15
Percentage growth of Linux new license revenue shipments over the last year: 28
Concurrent decline in new license revenue shipments for UNIX: 25
Number of Zumiez stores installing Linux desktops: 100
Estimated $-per-desktop savings over Microsoft alternatives at the Zumiez stores: 500
Scott McNealy dropped the first shoe at LinuxWorld Expo in August 2002. In his keynote address, Sun Microsystems' President and CEO said the company would be announcing a new Linux desktop at its SunNetwork conference in September. When the second shoe dropped at SunNetwork, it didn't appear to match the first. Rather than yet another Linux box, what the company announced was a desktop strategy meant to take advantage of the huge enterprise market for inexpensive and flexible no-name x86 PCs, and for cost savings in general.
The code name for the strategy is Mad Hatter. Here is how Curtis Sasaki, Sun's vice president of engineering, desktop solutions, explains it:
What we're announcing is a complete package, not just a box. You get a combination of hardware, software, services and back-end middleware as well. The hardware is a desktop with a Linux kernel, GNOME GUI, integration of the Java 2 platform, Mozilla, StarOffice and the Evolution application suite from Ximian. The differentiating factor is integration. What you get with one of our boxes is enterprise-ready and scalable, with directory, calendar, messaging server and Java Cards for access control as well. [Java Cards are Java-enabled smart cards.]
Target customers are cost-conscious companies with large populations of “transaction workers”. But rather than selling droneware for cubicle hives, Mad Hatter's angle is all about individual identity.
“Today CIOs want to know exactly what it costs per user to have e-mail, calendar and a directory account—and what it costs for security”, Sasaki says. What Sun wants is for the enterprise to populate itself with Linux PCs that are personalized at authentication by the user's Java Card, and there appears to be a demand for this. Sasaki says:
While customers are extremely interested in aggressively priced systems based on open standards, they are also interested in the most unique hardware aspect of this offering: the Java Card that allows the administrator to provision web sites and applications-based user authentication.
This brings us to a whole new classification and an acronym to go with it. “It's not a PC, it's an IdC—an identity computer”, Sasaki says. “Identity is a big deal. It's about getting access to your desktop no matter where you are, based on your credentials.”
Eric Norlin, an analyst at Digital ID World, provides some context:
Corporate IT has become almost a pure cost center. Just about the only IT efforts actually saving money (while increasing privacy and security) are in identity management. You can expect identity to move forward while other services tread water, because identity has the real promise of converting IT from a cost center to a profit center. With Digital ID you have a real possibility for ROI on IT investments.
Sun's intent also is to make a single identity work outside and between companies as well as inside the user's company. Sasaki explains how:
Liberty Alliance has an open spec developed by 115 companies from many industries. That spec answers the challenge of creating a way for users to sign on once for multiple services. You're going to see a lot of different Liberty-enabled services being able to utilize your identity securely. When there is a business relationship with an enterprise that also deploys Liberty-enabled identity—say, United Airlines or American Express—then you can actually move from one to another without re-creating your identity.
Why now? According to Sasaki, the Linux desktop software stack is finally complete:
In the last 12-18 months, LOTD (Linux on the Desktop) technology has really matured. A year ago it wasn't real. We couldn't deliver a complete desktop solution. Our office suite wasn't there. GNOME and KDE weren't mature enough. Mozilla wasn't at 1.0. Now GNOME 2.0 is pretty cool. StarOffice 6 is getting great traction. You've got a pretty nice product in Evolution. Now we're ready.
Sun will be putting together the first prototypes at its iForce centers before the end of the year and expects the first IdCs to start shipping in the first quarter of 2003.