I'm gonna be writing about Google's AppEngine for a while as it strikes me as one of the most interesting moves in the evolution of the web-as-platform that we've seen in quite a while. So let me get the boring stuff out of the way first...
A couple of years ago when Tabblo was a struggling startup in need of any and all distribution, I visited a senior executive at Yahoo who supposedly had a nose for new and interesting personal publishing applications. Our hour together quickly devolved into him trying to convince me that we should abandon our own user authentication system for Yahoo's recently announced browser based authentication. As he pitched and I listened all I could think was: yeah right, and what happens when one of your competitors is interested in doing a deal with us?
These things shouldn't matter when considering platform-level servicesâ after all, very little of Tabblo's special sauce came from the way we authenticated users (except for maybe some of those "special" security holes). The reality though is that when what you're building is intertwined with a emergent business model, these platform decisions are just as much politically driven as they are technically motivated.
And as Google joins Amazon in racing to provide web developers with critical bits of cloud infrastructure, we're about to see politics come to the fore again. Fifteen years ago, the question was: are you writing for Mac or PC? Today it is: are you writing in Rails, PHP, or Django? 5 years from now it will be: are you on Amazon's cloud or Google's?
Dave Winer was the first to point out that Google's move to open its platform to developers was a smart tactic to grease the skids for acquired startupsâ after all, if you are already running on the Google cloud, assimilation post acquisition is an easy task. And trust me, on a week when we're struggling to move just the hardware from Boston to Austin to get more compliant with HP's IT infrastructure, I can really appreciate this point.
But ultimately I think it's wrong if only because it seems to me that Google's acquisitive binge may be coming to a hiatus of indeterminate length due to maturity of its own business and greater macroeconomic forces. When looked at it from this perspective, AppEngine makes even more sense: why buy the pageviews outright when you can just toll them? All of those great arguments about long-tail application notwithstanding, letting others take the product risk while they ride on the back of your platform seems to be a much better bet to make.
And by the way, as far as platforms go, this one feels like a winner to me, even after only 30 minutes of perusing the SDK and docs. Unless Amazon moves quickly to integrate its own stack, S3/EC2/SimpleDB is going to feel like DOS in Google's Windows world. The latter's integration of the stateless webapp model, its own super scalable persistence layer, and a number of great open source pieces of infrastructure (way to use Guido guys!) possesses an elegance which maybe only engineers will appreciate short-term, but which is likely to pay off in terms of the stuff folks are going to build on top of it.
Again, these guys seems to get being along the grain of the web.
Postnote: Dave Winer has a really nice short writeup on the significance of this move by Google.