Original URL: http://www.theregister.co.uk/2008/10/09/linux_at_17/

Linux at 17 - What Windows promised to be

You can't stop the port

By Timothy Prickett Morgan

Posted in Operating Systems, 9th October 2008 17:20 GMT

On October 5, 1991, the young man who would one day become the world's most famous programmer - and the brand name and poster boy for the open source software movement - sent a message to a newsgroup announcing the birth of what would become the Linux operating system.

You can read that original message that marks the birth of Linux as an open source project, posted by Linus Benedict Torvalds, on Google's archive of a newsgroup called comp.os.minix .

It is hard to guess how many programmers and system administrators have been educated by the Linux development project, but it forms the core of what so many experts and newbies believe in terms of what an operating system should have in it, how that code is created, and how the systems software stack that rides atop of it is created and maintained.

It is safe to say that many millions of IT experts have been affected, either directly or indirectly, by Linux and the open source software movement it unleashed on corporations. While academic and government institutions had long since supported open source software projects as well as the Unix open systems movement, it is Linux - first and foremost - that made open source a commercial idea and one that corporations could embrace.

Linux has come a long way in those intervening 17 years, which are a bit like dog years with respect to how computer technology (both hardware and software) moves at an accelerated pace compared to other technologies and areas of the economy. It is hard to say if Linux is middle-aged or not, since the successors to OS/360 are still around more than four decades later (in actual time), and the original Unix is almost as old.

Even the commercial implementations of Unix are three decades old, and commercial Windows servers became a reality in 1994, more than 14 years ago. Who is to say how long any of these platforms will be around in production environments, but the history of the computer industry suggests that legacy platforms linger longer than many expect but lose their potency in the market ahead of when many might have hoped. In many companies these days, the only two alternatives are Linux and Windows for new applications - and some day, far into the future that is hard to conceive, these will be legacy platforms too.

What Windows Promised to Be

But it's remarkable that an open source movement backed by a handful of commercial entities with very little marketing muscle - at least compared to the established proprietary and Unix operating system providers who made so much money in the 1980s and 1990s - could take on the data center and, more importantly, get the begrudging support of the very system sellers who had the most to lose if Linux took off. This is a testament to the powerful idea of a cross platform, open source operating system. Linux is what Unix should have been and wasn't.

Linux is what Windows had once promised to be - at least in terms of cross-platform support. In the wake of the PowerPC alliance from IBM, Apple, and Motorola in 1991, Microsoft made a commitment to support Windows NT 3.51 on PowerPC chips. Windows eventually added support for Digital's Alpha NEC's and SGI's MIPS chips. Workstation maker Intergraph ported Windows NT 3.51 to its Clipper chips and said it was creating a port to Sparc chips from Sun. Neither ports saw the light of day.

Windows NT 4.0, which came out in 1996, only supported nothing more than f32-bit x86, Alpha, and MIPS chips, and by the turn of the millennium, only x86 chips were supported. (Interestingly, the PowerPC alliance also lined up IBM's OS/2 and AIX Unixes - the OS/2 was never delivered - and even Sun Microsystems' SunOS Unix was slated for the PowerPC chips. IBM also ported its OS/400 minicomputer operating system to the 64-bit variants of PowerPC).

While Microsoft has expanded support to cover Itanium processors - mostly at the urging of Hewlett-Packard, Intel's Itanium development partner and the one with the most to gain from Windows-on-Itanium for its high-end Integrity servers - Microsoft has not made good on the initial cross-platform promises for Windows server. Microsoft has suffered from this, but not as much as Intel has been helped.

The beauty of Linux is this: You can't stop a port to a new architecture, even if you wanted to.

That is why Linux is the most fully supported operating system, in terms of architectures, in the history of computing. (The open source BSD Unixes do a pretty good job covering platforms, so a tip of the hat there). And the coverage that Linux offers is not just broad, but deep. There are so many different Linux projects, each doing their own part to get it running on different iron - often for personal, not economic, reasons.

If being open source helped establish Linux in academia - where students could play around with the guts of the code as well as learn open systems software concepts derived from the many flavors of Unix - it is the breadth and depth of support that has given Linux a home in the data center. The lower price tag for commercial support compared to proprietary and Unix alternatives didn't hurt the commercialization of Linux either.

From 1994, with the release of the Linux 1.0 kernel, up through now, with the development of Linux 2.6.27 kernel, Linux has grown as an operating system, moving from a desktop toy to a system that can span the largest SMP servers in the world as well as becoming the de facto standard for parallel supercomputing, knocking Unix completely off its throne in less than a decade.

It took a while for Linux to get scalability and performance, but get there it did, and with the commercialized variants of Linux available from Red Hat and Novell, Linux is - and has been for a few years now - an absolutely safe choice. No one is going to get fired for buying Linux in the data center.

Linux Is Nice. But Benedictix is Better

But Linux does have its challenges. Growth for Linux servers has stalled, and this is because the paradigm for computing that is represented by Unix and Linux is unknown to many people, particularly small and medium businesses that adopt Windows reflectively. They go with what they know on their desktops, which is why Windows, not Linux, has dominant revenue share and most of the shipment share for server sales, quarter after quarter. Linux has grown to represent about a fifth of the revenue base, which is quite an accomplishment, but Linux has really only partially filled in the gap as Unix declined and Windows ascended.

In the past decade, Linux adoption was hampered by technical and support issues -it didn't scale and commercial-grade support took a while to bring to market. But in the next decade, the issues facing Linux are more cultural and inertial. It will take a tremendous force - perhaps a global recession or depression - to make companies think about learning Linux and investing in the technology rather than go with what feels like the safe choice of using Windows.

With the bulk of applications in the world certified to run on Windows, and Microsoft having some 800,000 partners, momentum is clearly with Windows. But with Linux on the desktop getting to be more practical with every passing year and with Microsoft contemplating a non-Windows future with its Project Midori operating system, Linux may yet get a shot at the title for operating system heavyweight. One this is for sure: There is no other contender for the title, no matter how much Sun Microsystems believes in OpenSolaris.

One last thing: Linux might have been the obvious name for the operating system, but Benedictix would have been more accurate. ®