all of its users to download the firmware patch, even if they don’t want to, or are intimidated by the process? Not all iPhone 3G users, after all, may be technologically savvy.

That’s not Apple’s only problem with the iPhone 3G launch. Its .Mac service—renamed MobileMe—has proven to be erratic and unreliable. Many third-party applications are unstable and buggy. Apple’s advertising, according to U.K. regulators, is misleading, because Apple claims that the iPhone 3G can access “all parts of the Internet,” but it doesn’t handle websites built with Flash or JavaScript. And in the United States, one customer has threatened a class-action lawsuit because the iPhone 3G is sometimes slower than promised.

The good news, of course, is that customers are still queuing up to buy the iPhone 3G. Why? Because inconvenient facts can’t stop a social phenomenon.

As Vaporous as
a Computing Cloud

Nephologists might define a cloud as a visible mass of droplets or ice crystals floating in the atmosphere. That’s helpful for cloud scientists. Computer scientists, however, view clouds as invisible masses of services floating in the Internet. Like meteorological clouds, compute clouds can be bright and sunny. They can also go dark at any time.

As discussed at length in

this issue’s “Business” column (see page 11), cloud computing is vulnerable to reliability problems. Apple’s .Mac cloud services, now called MobileMe, have been up and down since their July relaunch, particularly when it comes to email delivery. Ultimately, Apple offered MobileMe customers two months of free service as compensation. Google, too, has suffered reliability problems on its Gmail service, which went down in mid-August, affecting both its free and paid Google Apps Premium customers.

The cloud means more than e-mail, obviously. Amazon.com’s Simple Storage Service (S3)—a set of APIs for Internet-based storage, popular with website operators—went down on July 20, and stayed down for an entire business day. While Amazon kindly issued its customers an automatic billing adjustment, it was little consolation for websites that were down, or barely functional, while their owners wrung their hands in frustration waiting for Amazon S3 to go back up.

The message: The cloud is beautiful and seductive. The economics are attractive, especially since cloud-based services require only small monthly fees instead of costly hardware and software purchases. And they scale well. However, they are utilities, just like your power or water or broadband: When they go down, there’s nothing you can do until they go up. The good news is that reports of outages are rare.

The bad news is that outages happen at all.

VMWare’s ESX
Bomb Shows License
Vulnerabilities

When you use a cloud-based computing service, you understand that some days it might work better than others. When you install software on your own servers, however, you expect the software to keep working every day unless you do something to mess it up.

VMware is the poster child for an unexpected form of software failure: A software crash caused by poorly written license code. While the crash brought down the company’s ESX virtualization servers, this should not be seen as a specific vulnerability of virtualization. Instead, the failure should be viewed as a warning against any licenses that have the power to automatically turn the software off: a kill switch, you might call it.

On Tuesday August 12, users of VMware ESX/ESXi 3. 5 release 2 discovered that their licenses were incorrectly determined to have expired. Virtual machines that were running kept running, but any VMs that were stopped would not start up. That’s bad news for anyone, of course, who needed to reboot a virtual server.

To VMware’s credit, the company jumped right onto the issue, releasing a patch in less than 25 hours. Given VMware’s historic reputation for quality, it’s not fair to single them out for too much