Apparently Thomas Watson of IBM never actually said in 1943 that the world market only had room for 5 computers. Still, the misattribution’s been favourite fodder for years on lists of short-sighted predictions, along with Bill G’s equally misattributed “Nobody needs more than 640K of memory”.
The funny thing about history is how we’re now at a point where people are actually regarding that first nonquote with fresh regard. Sun’s John Gage, one of their original employees, once famously said that the network is the computer, and in this regard Watson’s nonquote starts to make some sense. To be more specific, substitute “network” with “distributed computing platform”. The idea is simple. Only a few companies have the resources and expertise to maintain an international-scale computing environment that applications can scale across to meet the gigantic range of demand the internet can provide. It’s also the source of a compelling business model to the potential owners of such “computers”.
Amazon Web Services have been the biggest and most prominent push in this direction for some time. Sun did come up with their Sun Grid product, but it was a dud by most accounts. Why? Because it wasn’t really connected to the internet. AWS (by which I primarily mean the EC2 computing services and the S3 storage service) are oriented all around supporting web applications and rich internet applications. They recognised the value in providing a service that small developers can build on with a reasonable expectation that should they hit the ball out of the park, that they’ll be able to handle any surge in traffic without going into the red ink for three years to come.
It was always strange that the world’s most famous distributed computing platform, Google, not be a fore-runner in this game. But that’s changed now. They’ve arrived with a flash and a bang. Scoble has videos of the launch, but the bare facts seem pretty cool. A Python environment with access to a storage service based on BigTable, and free accounts. The accounts are limited to 500MB of storage, 200 million megacycles/day CPU time, and 10 GB/day bandwidth, with the obvious business plan being to provide scaling resources beyond that for a fee.
Potentially it’s a huge announcement for web developers, and for Python. Google has a very strong brand in when it comes to reputed distributed computing power, and fears of platform lock-in are mostly eroded by the open tools architecture. It would require a rewrite to move an app from the GAE to your own servers (unless you thought about everything closely up front), but it wouldn’t be a huge one. It’s WSGI compliant, and it even comes with Django built-in. The only really unique part about the platform is their GSQL language, which is an acceptable change from the norm since BigTable isn’t a row-oriented database. That and some other features like Google Accounts (which they should really hurry up and turn into an OpenID service) integration, which is of secondary value to the average developer.
It’s limited availability and the first 10000 accounts have already been snapped up (and I missed out :-(), but there’s an SDK available for playing around with. Expect some nice experimental web apps in the next few months, afforded by the very low barrier of entry on this.