Death of the OS, re-birth of the kernel

I was walking through an insurance agency the other day. I figured -- if there is any place to see a lot of client apps running on the desktop, this would be it. A big server in a back room closet, Windows XP running Word and IE, a lot of people trying to talk someone through a claim.

As I glanced at each workstation, I noticed a common theme. One person had Firefox running, obviously surfing the Web. Another had IE running, but from I could tell, there weren't any other apps. On the screen, there was some sort of Web application that looked like it was connected to a remote server. (I was with the owner, who gave me permission to see how employees do their jobs) A few more agents were running Firefox as well, with e-mail and -- big surprise -- FaceBook up on another. As a small agency, they don't lock down their standard desktop with only IE and Word.

It got me thinking about why we need the traditional model of the operating system anymore. Yesterday, I posted about stand-alone devices that use a kernel (the system that runs basic I/O, memory management, etc.) as a thin client connected to the Internet. The reaction was very negative: we will always have operating systems, most commenters said. It's required to run the hardware.

It's my view that the Web has outpaced the traditional OS -- we have reached a point of no return for running bloatware, most of us need nothing more than a browser to do our jobs.

Companies like Oracle have long figured out that the Web provides the best distribution model for software. Meanwhile, in a future scenario where thin clients connect to the Net, the processor will be just as valuable because it will be leveraged for browsing and local disk access. Granted, high-end PCs will likely need to continue using high-end CPUs and some sort of light OS (although in the far future I see the OS as being customized to the activity, such as creating brain maps or as a workstation running Avid or even Apple programs for video editing chores).

Okay, so most of the readers here disagreed with the idea, saying I was defining terms wrong, etc. But I think it's fairly obvious what we all mean by an OS: it's Windows XP, it's Ubuntu, it's Mac OS X. It's the sprawling, complex software we use to run applications locally. No one shows you a digital picture frame and says, "and the OS is really cool" or that they can install other apps and run a virus checker (because, they can't). They show you the picture frame and say, look at how easy it is to view photos, and look how cute my kids are. Maybe you call that an OS (you have to admit it is a long, long way from Windows XP), I call it an app running on a Linux kernel, with very minimal memory management.

As I said, in 10-15 years, the OS will change. The definition of a computer will change -- it will become less about Windows or Mac or Linux and more about the Web. Simply stated: the OS will be reduced to what it should have been all along, which is a kernel to manage the system (and load a browser, hopefully something by Mozilla). All of the other components inside a PC have now caught up. We need "dumb" terminals again, connected to the Web this time, but lightening fast.

Now, a few posters brought up the Sun analogy -- this is a company that has been championing the cause of netbook computing for years. It's an old story. A decade ago or more, Internet connections ran slow, client computing was popular, and Web apps were few and far between. Today, broadband is getting faster, and Web apps are increasingly more powerful and feature-rich.

In the insurance agency I visited, thin clients would work just fine. There's no need for Outlook, or Word, or any other client apps. I'd say the vast majority of knowledge workers in IT could get along just fine with a thin client today. The myth that broadband is not fast enough is crazy. If it is not fast enough, why does everyone spend a majority of their time on the Web? If Web apps are so crude and un-useful, why are MySpace and FaceBook, Gmail and Yahoo Mail, and Salesforce.com some of the most visited sites on the Internet? And by the way, in the mid-90s, when I saw reports of corporate Internet use, it was rare to have someone spend all day online. We were annoyed when we saw someone who was online for more than an hour. Today, what's the first thing most people in companies do when they get into work? Fire up the browser. They check Webmail, visit their favorite sites, and then tap in to a corporate application -- using a browser. When they do use the OS, it is as a distraction -- they maybe play a game, or change their desktop wallpaper. This was not true just a few years ago, when most knowledge workers had a client for corporate apps and Microsoft Office running. An idea that was not ripe long ago can still become ripe, and just because thin clients failed the first time doesn't mean they will fail again.

I also wonder: why all the resistance? (It's futile!) Part of me thinks it's because we've all been raised up in IT to believe in the traditional OS model of client-server computing, it is what we know, and what we are paid to know. My challenge is to think differently about computers. The consumer mindset is changing. They want connected devices, not operating systems. Windows 7 may be a stripped down version of Vista, and Linux may become more popular in the enterprise next year, but the real long-term vision for computing has to be: reduce the OS to its roots, and make the Web the thickest pipe.

Next up: I'm going to find a few more examples of several more devices that, according to the most common and understood definition of an OS, do not have an OS.