SOA at rock bottom
It was bound to happen. About 18 months ago, SOA was all the rage. Every serious enterprise software vendor had to have an SOA strategy. Every enterprise customer rushed ahead with their SOA plans. Now Web 2.0 is soaring high while SOA — accompanied by SOAP and the whole WS-* stack — is down in the doldrums:
- "SOA may have meant something once but it’s just vendor bulls*** now," says XML co-founder and Sun blogger Tim Bray.
- There's growing frustration with the "relatively limited business impact achieved by SOA deployments" writes the influential John Hagel.
- "This SOA thing has been ongoing for the better part of 5 years and we still don’t have substantive end user benefits to show for it," says SAP's Jeff Nolan.
- "Where's the revolution? And the alignment? ... we're building the Titanic," opines VC blogger Vinnie Mirchandani.
When everyone is lining up to trash anything as comprehensively as this, you know it must be reaching some sort of bottom. We haven't yet reached the point where vendors start to renounce their commitment to SOA, but it can only be a matter of time. This is what Gartner terms the 'trough of disillusionment' in its technology hype cycle. If SOA were a quoted equity, it would now be at least 90% down from its peak valuation. Everyone is bailing out.
Interestingly, this is happening at just the same moment as early adopters of SOA are beginning to report dramatic, quantified benefits. As Joe McKendrick writes, IBM has reduced by three quarters the number of applications it has to maintain internally, while Verizon Communications says it has slashed its IT budget by 50% after adopting SOA. But these are isolated examples, and they all have in common a single-minded focus on service reuse as a means to simplify historic application sprawl.
I think the underlying reason for the wider disillusion with SOA is that it has been adopted in a vacuum. Enterprises have built SOA infrastructures without articulating clear business objectives (as opposed to IT objectives such as increasing reuse and reducing application sprawl) — perhaps in the hope that business benefits would somehow magically result. Of course they haven't. The advantage that Web 2.0 has over SOA is that it can be business-led, simply because business users can adopt it over the heads of their IT colleagues (which John Hagel has very cannily now advised them to do) and see the results almost straight away.
As inevitably as SOA has fallen from grace, sooner or later it will bounce back. But in what form? When on-demand applications hit their trough of disillusion in mid-2002, the term ASP had become so despised it had to be jettisoned, and only in the past year has Software as a Service (SaaS) been resurrected and won popularity. I don't think SOA has reached the same degree of revulsion, but I think the term will quietly start to go away, and all those websites that rebranded themselves from WebServices[name] to SOA[name] will have to rebrand themselves all over again (which is why I named my own web services and SOA site Loosely Coupled. I saw this coming).
Web 2.0 is on a hype curve of its own, though, and it won't be a pretty sight when it, too, fllips over and begins the plunge toward its trough of disillusion. The unreal business models of a lot of Web 2.0 startups mean that people will get their fingers badly burnt. Long-term, enterprises may adopt a lot of Web 2.0 technologies and best practices, but don't expect it to happen under that name.
What Web 2.0 does well is to add the collaborative, human dimension at which business outcomes operate, and which SOA has largely lacked. What it doesn't do is provide reliable, trustworthy mechanisms for identity, governance, service level consistency and payment. Over the next few years Web 2.0 will assimilate methods of providing those missing attributes, effectively remaking SOA for the real world. A lot of the work that's already been done in building the WS-* stack will be adopted, but it's inevitable that some of it will be discarded simply because the assumptions on which the specifications were based will prove to have been wrong. As Tim O'Reilly told The Register in an interview published this week, "there’s almost always a premature effort to standardise ... they inevitably get a lot wrong."
It's been interesting in the midst of all this to see Nick Carr and a few others ponder what role SaaS might play in all this. I think the answer is encapsulated in Jason Kolb's insight that "the real point is, the software needs to be written as a service." The missing link that connects SOA to Web 2.0 is a services mentality. SOA, SaaS and Web 2.0 are all services architectures; I see them all as part of the same continuum, and I believe on-demand SaaS vendors will play a central role in bringing the convergence of SOA and Web 2.0 to life.