At the Cloud Computing Forum, InformationWeek asked a distinguished panel why it was necessary for every hypervisor vendor to launch its own virtual machine runtime format. If we can see the need to move workloads from one cloud to another, a common runtime format would simplify the process. What will it take, I asked, a user revolt?This question warranted it the "smother it with a wet blanket" treatment by the panel. Jeffrey Barr, senior manager of Amazon cloud computing solutions, responded first. I agree that Amazon with some justification came up with the Amazon Machine Image, its preferred AMI runtime format, since the Amazon Elastic Compute Cloud was an innovator on the scene. But still, you would think Barr would at least agree in theory that a shared format would be useful.

Barr responded that "another way to look at it is to think of AMI as an output file" from the user that the Amazon cloud knows how to deal and load up to the appropriate platform. Both Amazon itself, third-party vendors, and open source projects are supplying tools to generate this output file to help users take advantage of the EC2, he explained.

"I don't think we're holding back any genuine progress by not documenting the AMI format," he concluded.

Joseph Tobolski, Accenture's director of cloud computing, backed Barr up. "Jeff's point is perfectly valid. You've got to wait until the time is right to reconcile those different formats," he said in an interview.

Unspoken but perhaps assumed, I think, is the understanding that the vendors have the right to launch their hypervisors and use their preferred formats until the marketplace conducts a shakeout, then everybody can tell who the winners are and what formats deserve to be long-lived and "reconciled" or even collapsed into one shared format.

Right now, we have VMware's VMDK; Microsoft's VHD, also adopted by Citrix Systems; and open source Xen, adopted with variations by Virtual Iron, Sun Microsystems, and Oracle, among others. So we are once again going to use the measure of de facto standards -- who wins the shakeout, who gets adopted by the broadest base of users -- to set a common standard years after users could have adopted one upfront and gone on with their data center reorganizations.

The formats aren't that different and the differences don't bring a variety of virtues that allow users to choose among them. As Barr noted, they're undocumented as far as the average user is concerned. Leave it to the cognoscenti to sort this out.

Ah, that's what we did in all the previous waves of computing change that have swept the data center, inflicting inoperability barriers where they didn't need to be imposed. We're doing it again. The problem is so easy to solve that the leading hypervisors currently recognize competitor's incoming virtual machine files and convert them to their own. There's not much rocket science involved; it's merely vendor arbitrariness that needs to be solved.

Margaret Lewis, AMD's director of software strategy, defended the current practice as better than in the past. She was referring, I think, to Citrix aligning its format with Microsoft's VDH, Microsoft and Red Hat agreeing to support each other's operating systems as virtual machine guests, and the DMTF standards body's effort to establish an export format, OVF.

"We see our software partners working more cooperatively than they have in years. Agreements are being reached and alliances are being made..." with the incompatibility of runtime formats one of those issues that will have to work itself out over time.

I agree, more cooperation is being shown on the virtualization front than software vendors have shown in the past, but with such a simple remedy at hand, is it enough? Knowledgeable parties inside ongoing software concerns may have a disdain for those users, those small-minded individuals who just can't understand why things need to be done the way they are. But I for one say bring on those revolting users. After this gang, I'd like to hear from them.