This week, debate over that urban legend -– as applied to the software industry -– will be the central theme of Software 2007, an annual gathering of the folks who buy, sell, and run software companies. Specifically, it’s the question of whether consolidation is killing innovation.

Clearly, unless you’ve been living under a rock, it’s kind of hard to not notice that software industry consolidation has sharply accelerated since 2000. And according to conventional wisdom, when you have fewer voices there should be fewer new ideas.

But let’s first ask if anybody cares. Is there such a thing as too much innovation?

Ask any SAP customer about how much they look forward to upgrades. Of course, in most cases, version upgrades are not necessarily innovation. They simply augment (or detract) from existing functionality.

And then there’s innovation that in actuality is little more than feature creep. Case in point: Microsoft Word 2003. An upgrade from Word 2000 –- which we fondly remember because, for us, it worked just fine, thank you. Yet the 2003 upgrade added “smarter” formatting that was supposed to be innovative. But in actuality, it ended up making our life more complicated and less productive. Obviously there’s such a thing as half-baked innovation.

Too much innovation -– good or bad -– can be clearly disruptive. But obviously, were the march of technology to cease tomorrow, you’d never be able to solve that perplexing integration problem, or discover how to rationalize IT service management.

So let’s return to the main point: if innovation is a good thing, does consolidation stifle it?

Obviously, acquisition strategies where vendors are swallowed up for maintenance streams certainly validate conventional wisdom, witness CA’s practices during the Charles Wang era.

But when a vendor acquires a microscopic startup, it’s usually about mainstreaming, not killing innovation. For instance, when BEA bought 12-person SolarMetric about a year and a half ago, it was a shortcut to adding a critical piece of Enterprise Java Beans 3.0 object/relational mapping technology to its products.