–
by Jack Bryar –
Open Source Business –
Linux
Kernel
2.4.1 is out,
just in time for
LinuxWorld 2001. It’s a show that
has some software, but not as much as you might expect, and that’s
because of 2.4. It’s not just that Linus’s new kernel is — by his count
—
15 months late. It is because, for the moment, shrink-wrap
style desktop applications have taken a back seat. The new kernel is no
next-gen desktop toy, but an enterprise tool (and that’s also why it’s
late). The next question is, can enterprise developers play by Open
Source rules?Developer shows tend to feature a fair amount of hardware, and often
lots of new software. LinuxWorld 2001 has a fair share of both, but not
as much software as one might expect.
It’s not surprising. After all, the Linus Torvalds and his team
didn’t officially release 2.4 until a few weeks ago. 2.4.1 has just
been
released. Although virtually all software developers have been tracking
the development of the new kernel over the last two years, that didn’t
mean they were ready to release product that took full advantage of the
kernel’s new features.
Development has been slow and painful. Torvalds, himself, has
admitted, ” the original plan was to try to aim for a nine-month
productization cycle.”
But Torvalds has said that much of the delay has been in response to
pressure to scale up Linux to service the back end requirements of Web
developers and the needs of enterprise managers. Torvalds said, “My
original main goal was to clean up the SMP scalability to four CPUs,
and it kind of grew into a major file-system redesign”.
Despite the fact that most of the work on 2.4 was done by Torvalds
and many of the same loose team of core developers that have worked on
Linux for the last decade, corporate influence was everywhere. Nearly
all contributed development and testbed support. And they complained
more — and developed more independent code. The final result reflects
that corporate effort; 2.4 is a more corporate-focused product.
In fact, for most desktop users Linux 2.4 may not even be that big
a deal. The support for USB and firewire devices is good, if overdue, and the support for PCMCIA is nice, but patches and code have been
available to address these issues.
The more dramatic changes were designed to meet the requirements of
back end and enterprise applications. The 2.4 kernel provides better,
more scalable symmetric multiprocessing (SMP) support, which is
essential for multiprocessor machine. The new Logical Volume Manager
allows for terabyte-sized files and lets storage volumes to be spread
over multiple disks — critical for enterprise applications. The number
of supported simultaneous processes is now limited only be the amount
of RAM in the system, and Linux can now support up to 64 gigs of
memory.
The kernel now officially supports a variety of 32- and 64-bit
architectures. Further, the kernel has been restructured and made far
more modular, allowing third-part developers to patch, shave down, or
extend the architecture to accommodate mainframes or handheld devices
as needed.
This is not just the operating system of desktop hobbyists anymore. That’s just
fine with most industry evangelists, especially hardware vendors.
People like Intel Vice President William Swope, told Computer Reseller News that he saw Linux’s future less on the desktop and far more as the guts of Intel-centric back-end processing data center
environments. His view is widely shared. IBM’s Linux champion, Dr. Irving Wladawsky-Berger agrees
that Linux belongs in the center of IP network environments ranging
from straight Web servers to more sophisticated “added value” network
settings and in a variety of emerging server-side applications
environments.
The problem was — the code was late. Sources say that Red Hat and
other Linux distributors had their products ready to go in the fourth
quarter of 2000, and effectively got stranded. When Torvalds finally
declared Linux 2.4 “ready” only Red Hat was shipped an upgrade (using
the “preproduction” 2.4 kernel). Other vendors were less prepared. In
any case, the delay blew up any end-of-year purchases by corporate
customers.
There has been plenty of hidden (and not so hidden) irritation among
corporate types with the delays in getting 2.4 finally, finally, (no
this time we really mean it) finally mature and ready for general
distribution. As Swope said in an interview with VAR business, “Time
to market is time to market, and it matters.It [would have been] nice to
have the software earlier.” But, as IBM’s
Wladawsky-Berger said to
TechWeb’s Barbara Darrow, “It’s finally here. Now we can move
on.”
But even more important than a timely release may have been a
well-coordinated one.
Critics outside the Linux community such as MIT management professor
Michael Cusumano think that coordination is a big problem. He suggests
that Open Source developers inevitably violate many key
principles of software development, making it difficult for
commercial developers of applications to develop products because “they
can’t control schedules.” Torvalds has admitted that, “To some degree,
I
got ‘what do you think the timing will be?’ kinds of questions,” from
vendors, although he insisted that no one pressured him to release the
code early.
But a lack of coordination means that developers cannot integrate
their offerings in a timely manner. They cannot coordinate releases,
publicity and marketing events in a coordinated fashion. More than one
system integrator I talked to suggested it was time to
“professionalize” the development process, although none would say so for the record.
That’s unlikely to happen. The cult status Torvalds enjoys
is not based on hype but on fact. At present, Torvalds and his
kernel represent the one standard in an amorphous codebase that lacks
much cohesion. Want a GUI? Try two or three. Or build your own. Want a
journaling file system. Take your pick. Need a properly working
browser?
Good luck. In fact, if you need anything at all in the way of a utility
or application, it’s likely that many someones have taken a run at
building it. That’s part of the problem — there’s so many applications,
half built applications and, not-all-that-well-built applications
floating around that that trying to find appropriate code for a
particular application can be like trying to find a particular
snowflake in a blizzard. In the midst of this, Torvalds and his kernel (and his
development process!) have been the one constant.
And despite all the rhetoric about “choices” most developers crave
standards. Linux is attractive to both developers of embedded systems
and mainframe based solutions for three reasons: It’s documented, it’s
free, and it’s a standard that developers can trust.
Besides, the most remarkable thing about 2.4 is how well it’s
configured. I admit earlier versions of Linux always looked like a mess
to me. But 2.4 is clean, modular, and surprisingly logical code — a
model for commercial developers. It should be far easier for even an
average programmer to “adapt and extend” this new kernel to a variety
of emerging applications and devices. As a result Linux and Linus are more essential than ever to developers of products ranging from next
generation telecommunications equipment to advanced-but-inexpensive
home electronics equipment.
2.4 was late. And most of these products aren’t part of LinuxWorld
2001.
But they’re coming.
NewsForge editors read and respond to comments posted on our discussion page.
