Open Sauce Software

Intrepid Ibex, the 8.10 release of Ubuntu due this Thursday, is not a milestone for servers, but it comes at a time when Canonical, the commercial backer of the operating system, is starting to emphasise the server side of its business.

What server side, you may be asking. with Ubuntu on about 60 percent of Linux desktops, it's easy to forget that the operating system has a server version, in the face of strong marketing from Red Hat and the others. But the server is there, and the company says it's strategic.

"I don't think it will be possible to make a lot of money, or maybe any money, selling the desktop," Canonical CEO Mark Shuttleworth said at the Ibex announcement on Monday: "We're not going to try to make money selling the desktop." The company's revenue at the moment comes from support and custom engineering, and there's more potential for that sort of revenue around servers.

Ibex adds some interesting features to the server version, including the ability to build virtual machines quicker, and better support for virtualisation under Xen and KVM. It also has improved RAID and an easier firewall.

But Canonical doesn't expect a flood of users to switch production systems across. This version will only have 18 months maintenance, while Hardy Heron (8.04) is the Long Term Support edition, guaranteed to be looked after for another four and a half years. This time around, most people will treat the new features as technology to play with.

So how well is Ubuntu doing on servers? The OS has maybe a five percent share of Linux servers, Canonical's marketing director, Andrew Rodaway, told me last week. That's subject to the usual bucketful of caveats with open source figures. It's virtually impossible to tell the market share of a free product.

But is he being modest? At ZDnet in May, we found that around 19 percent of Linux server users had Ubuntu, in a reader survey. The same survey found that fifty percent of respondents had Linux.

This figure, of course, is based on only 195 response, from ZDNet UK readers, so it's not a true picture of the whole world. It also counts user sites, not servers, so the result means that 19 percent of the people who responded had Ubuntu in their server room somewhere. They might well be testing it on one server, while their production systems are all on Red Hat or Suse - or Windows.

I get the feeling that Ubuntu server has a bigger mind share than its estimated share suggests. In which case, I'd expect to see a lot of people trying out the new features, ready for the next Long Term Support release.

A year on from a spectacular controversy in quantum computing, it looks like rationality is breaking out.

Last year, D-Wave Systems, the only commercial start-up in the highly demonstrated a 16 qubit quantum computer. Or not, as the case may be.

The problem was that quantum computing is still at an early stage in the labs. It's rare to get two qubits, and to persuade them to exist long enough to do anything. D-Wave claimed its "adiabatic" approach got round the problem and performed a demo. But it was effectively a demo on a "black box" - and was criticised by academic quantum computing specialists, because D-Wave did not actually do the measurements which would have proved whether it really WAS a quantum computer or not.

Scott Aaronson, whose marvellous Shtetl-Optimised blog is required and delicious nerd reading, said D-Wave's Orion system was "as useful as a roast-beef sandwich". This didn't stop D-Wave from promising great things - earlier in 2007, it was talking about having a 1024-qubit system available by the end of 2008.

One year on, things seems a bit less confrontational. Geordie Rose - maverick founder of D-Wave - actually visited MIT, to talk with Aaronson, and Seth Lloyd, the co-inventor of the theory of adiabatic quantum computing. There are write-ups in MIT Technology Review from Lloyd and Aaronson. Aaronson says, in his blog: "The people at D-Wave are not conscious frauds; they genuinely believe in what they’re doing. On the other hand, much of the publicity surrounding D-Wave can be safely rejected."

For its part, D-Wave seems to be turning down the hype. It does expect to manufacture its next step - a claimed 128-qubit quantum computer sometime in the next couple of weeks (see Geordie Rose's blog), still way beyond anything in the peer-reviewed literature, but a step down from the 1024 qubits we were promised last year.

And Rose is now very circumspect, explaining to me that there are no plans for a big demo, and that this is still very much a prototype: "No demonstration is planned and hopefully there are no such reports. We are about to go to glass with the first rev of a 128-qubit design."

And meanwhile, if quantum computing is still a bit opaque to you - Scott Aaronson has written an article for Scientific American on the limits of quantum computing. I've yet to get hold of the final published version but Aaronson's early draft is beautiful.

I didn't come away from the Smartphone show this week much wiser about the prospects for open source phones. Android is looming and Symbian is soon-to-be-open. But I did gather one thing: the main competition is dumb-phones.

Symbian is going to be open source, the company's research chief David Wood told me again at the show. But It can't happen immediately. In fact, even the first incremental step can't happen till Nokia's purchase of Sybmain goes through and the Symbian Foundation is set up - sometime round the middle of next year. At that stage, a small amount of the code will go open source, and the rest of it is available under a flat fee to Foundation members.

There's going to be a complex procedure in turning code over to the open source community - so the amount of open source code will ramp up gradually over the following few years. This bit is going to be complicated - if you don't believe me, read my colleague David Meyere's interv iew with Wood.

But the big picture is that Symbian will be vastly cheaper to handset makers and OEMs. It goes from today's licence fee per handset, to next year's Foundation membership fee - and eventually, when it's all open sourced, to nothing.

That's more or less what Symbian has always promised - the company wants to get the OS on bog-standard feature phones (somewhat of an uphill struggle so far) and the only way to do that is to make it dead cheap.

Meanwhile, Android is starting out as an open source platform, so handset makers have yet another avenue for cheap devices. It's just a matter of which OS the vendors choose. Or is it?

Not necessarily, according to Charlotta Falvin, CEO of The Astonishing Tribe - a user interface specialist with software in millions of phones: "The costs of switching platform are huge," she said. "If people hesitatie to make investments, they may stay on existing platforms."

In other words, as the recession bites, operators and handset makers selling to the low end of the market will stick with existing closed "featurephone" platforms, and not move to the open platforms typical of smartphones.

Instead of putting money into a platofmr change, they will just pimp their existing platforms, apparently, and try to make them look new.. That's something Falvin is very happy with - it's the stock-in-trade of her Tribe.

"Last year, operators wanted an iPhone-killer," she said. "This year, they want to make phones look more expensive."

The whole thing means more fragmentation, not less, she said, and it sounds a plausible view to me.

A year after buying XenSource, Citrix has evolved its marketing strategy for the Xen open source hypervisor.

Virtualisation will be more about application delivery in the long term, and rival VMware will be limited by its server consolidation focus, Citrix experts told me on a briefing day last week, at the company's UK headquarters.

I had detailed briefings on Xen from Simon Crosby, Citrix CTO and founder of XenSource, and a meeting with Ian Pratt, the Cambridge professor who manages the Xen.org project and is also director of future products at Citrix.

All this took place in an 18th century mansion that also featured in the James Bond movie Thunderball. We were lectured in the room where he was almost killed by a physiotherapy machine - and found the experience only slightly less stretching.

The application delivery part is no surprise - that's the focus Citrix had before it absorbed Xensource. But that, and the fact that Xen is open source, will trump rival VMware said Crosby.

VMware is forced to keep value in the the hypervisor, Crosby argues, and in the job of server consolidation. That pushes VMware to try and annex the server management role, with specialist virtualisation technicians doing a job that should be done better by products like Tivoli or OpenView.

Being open source, Xen's impact on the market - and its market share - is very hard to establish. Crosby and Pratt both spoke of large numbers of "cloud" providers building business systems on the free version of Xen.

Overall, they argue that a strong story around application delivery to the client will make Xen indispensable in the server room - and the cloud part of the story suits it to the future evolution of IT.

Apps are supposedly migrating to third party cloud providers - if the IT department and the cloud provider both use Xen, this could be a matter of just moving a VM across.

There's plenty more to say onb this subject, but it was a very interesting. day.

Here's some viral marketing - ironically enough for an anti-virus company. Norton Fighter is a Japanese superhero, saving the world from viruses in black body-stockings.



Strangely, this has taken at least a year to reach me, despite its obvious excellence, so I apologise if I'm the last to appreciate it. It ends on a cliff-hanger, but Part 2 is easy to find on YouTube.

It looks very Power Rangers to me, but online afficionados say it's more Masked Rider, and bat around terms like Sentai. Norton Fighter is also available on YouTube, as an Akihabara street play: