Open source's integration problem (?)
It's true that open-source applications and infrastructure don't necessarily play well together today, as Microsoft suggests. But it's also true that more than enough integration is in place for current market demand.
Glyn Moody has an excellent article in Redmond Magazine on open source and interoperability. As it turns out, sometimes it takes Microsoft to notify the open-source community that for all the great things we've done, we sometimes fall short. One area that open source had traditionally failed in was in stitching together an end-to-end solution, as Nick McGrath (a friend and a wonderful person) suggests:
Of all the accusations Microsoft has leveled over the years against open source, perhaps the least contentious is that it lacks the tight integration offered by Microsoft's own products. As Nick McGrath, director of platform strategy for Microsoft in the United Kingdom, puts it: "One of the problems I've seen with open-source software is it doesn't take on board some of the issues that customers have around interoperability and integration. Open-source projects tend to offer a very specific point solution."
This statement has become less true over time, though it's still the case that there is no one open-source vendor providing seamless interoperability between disparate pieces of enterprise software (in the way that Oracle and Microsoft do or attempt to do). My bet is on Red Hat to become that company over time, but in the meantime, we're not there yet.
However, as Glyn suggests, LAMP, the Linux Foundation, Eclipse, and various other open-source organizations have arisen to help resolve incompatibilities. Even so, let's be candid: we're not yet at a point in the industry that we need to have seamless integration between Pentaho and OpenBravo. We're just not. We are definitely at the point that customers want an integrated LAMP stack, but usually if they're putting an open-source application on top, they're looking to that application vendor to take care of the synthesis of the pieces below (and we do).
In fact, it is generally integration into existing IT infrastructure that matters most today. I've never once been asked if Alfresco integrates well with Zimbra or SugarCRM, though all three companies are doing exceptionally well. Why? Because demand for open-source applications has not yet centralized within enterprises. Most initial adoption still happens at the departmental level and spreads from there. So, for now we're getting asked to integrate well with Red Hat/SUSE and MySQL (+Tomcat or JBoss or Liferay, depending on the application), or with SAP or Oracle or Microsoft.
The next five years will be all about integrating open-source applications and infrastructure with incumbent, proprietary systems. Five years out, enterprise traction will be such that technical integration and corporate mergers between open-source applications will make sense. We're not there yet, but it's coming.