This week no fewer than four people I respect in the cloud ecosystem cautioned me not to get to enraptured with cloud APIs. The argument was similar across their critiques: lots of talk about them, very little buying through them, they are interesting for geeks and obscure outside that circle.<p>I was shocked--but my shock in itself was alarming to me. Am I over rotated on the broad platform characteristics of cloud computing? Are cloud infrastructure services more valuable as a complete resource management applications than as modular horizontal platforms?<p>The argument against API centrality, as far as I can surmise goes something like this:<p><pre><code> Chill on the API love fest man. What really matters are the core functions of resource management, provisioning and user management. What's really new about the cloud isn't an API, we've had those, its the standardization and control of server and network resources en mass. Also, wake up, most people are going to consume whatever prescriptive GUI framework the cloud provider puts forth and be happy. Customers of XYZ big hosters are going to pay far more for a portal experience than an API.
</code></pre>
Here is my critique back:<p><pre><code> 1. A buyer, is not always just a buyer. Say 90% of cloud resource 'buyers' use a non API method of consuming resources. This doesn't refute the importance of the API. See reason #2.
2. Innovation happens elsewhere. Twitter knows this well. Do you use Twitter or do you use Tweetdeck on the Twitter platform? Assuming you follow the logic of the resource, network and user management argument above--it actually should create more incentives to open up an ecosystem of potential consumption and integration methods. The cost of creativity gap between a large organization, and a lean entrepreneur to create value ad above the core provisioning management system is vastly different. Twitter is a relatively simple distributed messaging system; the possibilities in infrastructure and data services are even greater.
3. Heroku, Engine Yard, JungleDisk, Storsimple; these are not just alternative management and control GUI's, they fundamentally ad another layer of real value. They are early examples of #2--and these are very early days.
4. If you are really pursuing a managed hosting offering where the customer is more interested in the service relationship than the platform automation and compatibility the API may definitely be less relevant. The managed hosting marketing is also orders of magnitude larger than the current cloud infrastructure market; however, none of that means the attributes of the managed hosting market = the cloud infrastructure market. They are juxtaposed but separate along key strategic axes.
5. Smart buyers with long term vision care about the API story for all of the above reasons. It is a necessary but not sufficient aspect of the offering to them--but again bad logic to confuse sufficiency with necessity.
</code></pre>
*<p>The question behind the question has to do with the strategic nature of cloud infrastructure services in the long term. Are they a corner case deployment, a hackers liberation sandbox with some interesting but limited mutations, or the fundamental platform of the future?<p>I have opinions there :)