It has been this change from one-time license fees to recurring support fees that forms the main structural change in the business. The provision of source code, open or not, is a secondary matter. I know that many will now be up in arms about this contention, so please consider a database example and imagine you come across a bug in the database, what do you do? If you have PostgreSQL, you could fix the source code directly. But given the complexities of the code base and the practicalities of either forking a new version or feeding back a bug fix, few people do this, and certainly no one would rely on it. Instead, you implement a workaround in your own code (albeit one that is probably facilitated by being able to see exactly what the database is doing via the source code). Now consider a parallel situation, but one in which you use Oracle. What you are meant to do is call up the support line and dutifully wait for a fix. But that could take years. Instead most people will also implement a workaround. That is preferable to wasting precious cycles educating a support rep dwelling deep in cubicle-land after which you have no choice but to wait and pray for that coveted but all-too-elusive return call. Practically speaking, both approaches net out to the same result.

In the past few years a number of open source projects have emerged that are undeniably superior to existing enterprise solutions. The Apache web server is a prime example. This superiority has changed the decision dynamic between commercial open source and modern enterprise software. Rarely is the decision now reduced to an open=good and closed=bad orthodoxy. Consequently, over the last 10 years, we've seen piece after piece of traditional enterprise infrastructure tumble to open source. I know that IBM, BEA, and Oracle, et al, will claim their infrastructure software offers greater performance and is more reliable or feature-rich. Unfortunately, a rigorous comparison proves that this claim often does not hold up. More important in most cases the vendor's argument in itself isn't relevant, since the open source equivalent is good enough. In addition there is no upfront cost and support costs are dramatically lower as well.

The bottom line is that I don't see a future for enterprise software infrastructure components.

Applications, though, are a different matter. The key difference between a piece of infrastructure software and an application is that an open source developer is a user of infrastructure and knows exactly what is needed, but developers typically have no clue about business application requirements. This would make me long on SAP and short on infrastructure software like BEA. Unfortunately for me, making a dime on this thesis may be hard because infrastructure vendors have been swallowed up through acquisitions. Such was the case with BEA by Oracle not long ago.

Enterprise software applications, however, face a threat from a different corner - SaaS. The differences between SaaS and on-premise software, whether open source or enterprise software, are more concrete. One way of thinking about SaaS is that you basically outsource the application in its entirety. But whether this is a good idea or not depends upon the precise nature of the application and its purpose in the business.

SaaS used to be referred to as utility computing. That remains an apt description because it makes the all-important point that a lot of processes, whether maintaining a customer database or an HR performance review process, are essentially commodities. There really is no IP or competitive edge in how you store someone's name and address. So it makes perfect sense to get out of IT support, needless capital expenditure, and high-risk implementations required by an on-premise solution, and instead just rent Salesforce.com or equivalent. The converse is also true. If you have a software-driven process that is part of your unique competitive edge (for example, how UPS routes their trucks), it makes no sense at all to outsource that application for two reasons. First it represents a company's competitive edge, which needs to be honed over time and so should be an area of ongoing investment; and second, there are no economies of scale that can be brought to bear on what amounts to a unique application. Of course today competitive edge applications like these are more likely to be built from open source components rather than a heavily customized application package, so ultimately this area of the business is by no means a sanctuary for enterprise software.

"Economies of scale" also brings in another consideration. A small business with 10 sales reps installing and running their own CRM system, instead of going with a SaaS solution, is something I would describe in charitable terms as "brave." With SaaS, you can piggy-back on a product, infrastructure, and support organization far beyond what any small or medium sized business could every even dream of. But how does this logic scale? What about organizations with not 10, but 10,000 sales reps? Situations like these need to be looked at on a case-by-case basis. Increasingly, large companies are moving toward SaaS systems for the more standardized processes, but are maintaining on-premise control for the more customized ones. On the one hand supporting 10,000 users requires a large amount of infrastructure that does not get shared for the simple reason that it all gets used up. That argues for going with an on-premise solution. On the other hand, SaaS systems do a better job of rolling out useful new features that present real benefits to the end users. Of course, you could argue that if IT is not your organization's core competency, you should not be the IT business, but I would argue that with 10,000 sales reps you need to be in the CRM business because with that kind of investment it is probably an integral part of the company's competitive differentiations.

Open source and SaaS are both revolutionary architecture changes that affect every business. The following figure gives a summary of my view of the world. But just as the client/server computing revolution eclipsed, but did not demolish the mainframe, I see the enterprise software model refocused into serving primarily the very largest organizations' competitive-edge application requirements and essentially removed from both the SMB and utility markets as well as the internal custom development markets by SaaS and open source, respectively.