Platforms, products and services

The key point in Greg Gianforte's article about platforms is this: "On demand eliminates the need to create a proprietary technology platform as a competitive differentiator." That universal truth applies just as much to Web 2.0 and to service-oriented architecture as it does to on-demand. All three phenomena are instances of the same fundamental shift in computing, away from vendor-specific platforms to a universally shared, open platform.

But vendors are so attached to the notion of owning platforms that they can't stop themselves trying. I wrote last week about the emergence of marketplace platforms in the on-demand/Web 2.0 arena. It's far worse in the SOA world, where every major vendor is pitching customers to build an enterprise services infrastructure using their own product family as a 'platform'. In doing so, they trample roughshod over the fundamental principle of standards-based SOA, which is to break free from vendor-specific architectures.

The only technology stack that matters in the Web 2.0 era will the one built using HTTP, XML and (optionally) assorted WS-* standards to provide a framework for publishing and subscribing to services. It is then up to the individual service operators how they implement those services. Done properly, a services architecture makes it irrevelant how services function internally. They are completely opaque black boxes. As Greg goes on to say, "the on demand model renders identity of the stack's individual components meaningless."

Conventional vendors (and indeed many of their customers) have a lot of trouble adjusting to this new regime because they have a product-centric mindset. They feel the technology features of the products are what differentiates them. But in a services-centric world, customers should have no interest at all in the internal workings of a service, so long as it meets the contractual commitments laid down in its service description (ie, it does what it promises to do). Being able to tap feature-rich software is irrelevant if it doesn't deliver the service they license it for.

The interesting consequence of adopting a black-box implementation is that, once providers are free to choose whatever internal platform they find does the job best, they most often choose open source components. RightNow Technologies itself uses MySQL, Linux and Apache (although it can also use Oracle and the rest for individual customers who choose to pay more for this option). As Greg says, "Open source commoditizes the stack."

There are quite a few high-level standards that still need to be fleshed out before the services marketplaces I mentioned previously can be equally standards-based. But that is how they will have to end up. Customers must pull back from getting drawn into discussion as to which marketplace has the most features or the best implementation. The first and only question to ask is, 'How well does it interoperate with other service providers?' If it doesn't, then sooner or later it will be overtaken by others that connect better to the Web 2.0 fabric.

The most crucial design pattern of Web 2.0 is the standards-based interoperability that creates a level playing field between competing services. This revolutionary change deposes vendors from their proprietary platforms and makes customers kings.