In other words, we seem to be coming from a push-based era of command-and-control management and are heading into an era where more and more work is being conducted using a decentralized pull-based model that’s more scalable, efficient, and leads to increasingly innovative outcomes. This also demonstrates how the network effect driven model of the consumer Web can actively remake the internal IT landscape as competitive entries in the form of SaaS applications, mashups, and enterprise cloud computing. Thus these new software models can not only co-exist but will thrive and even push out existing ‘IT approved’ applications.
Back in early 2006, I spent a lot of time thinking about the whole pub/sub, push/pull model, and how it would fit into where I though integration was going. In particular, I spent a long time chewing on Rohit Khare’s thesis on ARRESTED.
The shift we are seeing is not just in IT management, it’s also in data consumption.
I realized that the vast majority of integration scenarios I was trying to simplify were on the end user side – consumers want to pull data when needed, not have it pushed to them. They also want to mix data from multiple sources. And they want to do it on their terms. A push model requires someone to push, and that someone has to be centralized IT.
At the time, the shift in integration models seemed to parallel what I saw on the Web. So SnapLogic ended up with a REST architecture that balances flexibility in data access and transformation, self service, and enough access control and security to support enterprise installations.
Give the number of pull applications that have appeared in the past two years, I’m glad I didn’t decide to push.