Last week I wrote about the way specific technologies have lowered the barriers to entrepreneurship worldwide: Amazon Web Services, Google Drive (and Google Search), mobile devices combined with widespread wireless Internet service, and Facebook.

Most of these tools are what author Jonathan Zittrain would call “generative,” in that they are technologies which engender further innovation, thanks to their openness and extensibility. Like the earliest personal computers, AWS is an open-ended tool with virtually no restrictions. You can use a credit card to purchase an instance — a virtual server — and you can set it up and use it in almost any way your creativity will allow.

That creativity has enabled thousands if not millions of entrepreneurs to try out their ideas, build websites, and host data for mobile applications at a minimal startup cost.

In contrast to generative technologies are what Zittrain called “sterile” technologies: Black boxes that are locked by the manufacturer to a single set of uses, and which can’t easily be modified or extended by the customer. His example of a sterile technology was the iPhone, which Apple controlled completely at the time of its launch. That turned out to be a pretty poor assessment of the iPhone’s capabilities, because over the past seven years, since the launch of the App Store in July 2008, the iPhone has given rise to hundreds of thousands of independently developed apps, many of which have turned into businesses or even platforms of their own.

But the question of control is an important one. If you’re an entrepreneur, you don’t want to build your app (or service, or product) on top of a platform that is too controlled by its owner.

In a previous tech generation, companies that made DOS or Windows utilities often found themselves in this predicament. If their software addressed a shortcoming of the platform, or offered something too valuable, Microsoft might just add a feature to a future version of their operating system that rendered the utility obsolete, pulling the rug out from under the entrepreneurs’ feet.

Similarly, companies that have built large followings on Facebook have found themselves suddenly captive to the company’s new business model. It used to be possible to amass thousands or millions of “likes” on Facebook, and then use that channel to deliver marketing messages to your fan base or customers. No more: The organic reach of posts that brands put up on Facebook has dropped to almost zero. If you want to reach customers, you have to pay Facebook for distribution, even if you have millions of people who have liked your page. Bummer, especially if you were counting on getting that distribution for free.

Content publishers that grew to rely on inbound traffic from Google searches have been kneecapped when Google changed its search algorithms to penalize “content farms” and linkbait. That’s a good thing for most of us — unless you’re a content farm.

By and large, entrepreneurs are not concerned about this kind of risk. As long as Facebook reaches 1.44 billion people per month and almost a billion every day, it’s going to be an irresistible marketing platform. As long as AWS provides cheap, flexible, easy-to-use computing power, it’s going to be a smart choice. And as long as Google indexes billions of Web pages and remains the primary search tool on the Internet, you’re going to want to optimize your site for Google’s search engines.

If any one of those platforms changes its rules to the point where it becomes untenable, businesses built on top of it will have to adapt, just as they adapt to any other changes in market conditions.

But it does suggest that entrepreneurs and other tech decision makers would do well to pay more attention to the architecture and the business models of the platforms they’re using. The “generative or sterile” distinction may be too binary to be useful. Open platforms don’t necessarily have to be open source. But the question of relative openness is important.

Is the platform open enough that you can build what you want, and deploy it to the customers you want to reach, without being overly dependent on a gatekeeper?

Are the gatekeepers, if they exist, reliable and predictable enough that you can plan around them?

And what are the rules of the platform? Can you build something on top of it that will itself turn into a platform?

The answers to these questions may wind up being more important than you think. Because if you choose the wrong foundation to build on, you could wind up limiting yourself right from the start.