In a recent and provocative article called "SOA (the term) is dead, but SOA (the architecture) lives on" for SD Times David Rubinstein addressesthe opinion that while SOA has gained a lot of momentum as an architectural principle, it might be dead as a term. He quotes analyst Jason Bloomberg, who considers SOA as a bad term. In his opinion, SOA as a technology has already died due to Cloud Computing and due to the intrinsic complexity of Web services.
To prove his standpoint, Rubinstein quotes experts like Ian Goldsmith, vice president of product management at SOA Software, and Paul Fremantle, CTO and cofounder of WSO2, among a couple of others. Most experts, according to the article, believe that SOA has shifted from a technology stack that comprises standards like SOAP to an architectural principle. In addition, a majority of the experiences delivered to customers is out of the control of SOA vendors, which was different a few years ago and which is why Cloud Computing and other technologies have gained very high importance.
Goldberg from SOA Software said:
There was this pie in the sky about dynamic business processes, the ability to find a service to use and connecting to it automatically. That doesn’t happen. BPM and process automation are great, but it’s automation of relatively static processes. There is no dynamic discovery of new partners to integrate withy static processes. There is no dynamic discovery of new partners to integrate with.
The experts quoted in the article warned that the term SOA now implies several liabilities, for example:
- It has been over-promised,
- Some of the WS-* standards have fallen to the wayside.
On the other hand, SOA technologies such as REST, are considered fundamental. But the trend, has shifted away from services to APIs. Ken Godskind, vice-president of monitoring products at SmartBear is quoted as follows:
SOA is a category; APIs are a general category that fall under that. When I look at Web services, I see just XML RPCs over HTTP. Serving users in richer fashion will be the add-in. For applications to deliver rich functionality, it’ll be primarily with rich navigation and control containers making requests to Web services to pull in data via REST, SOAP and JSON.
What do others believe? It should be noted that the death of SOA as an implementation technology is not a particularly new topic. For instance, JP Morgenthal already discussed this issue in his own blog posting "SOA is Dead, Long Live Distributed Computing" in 2008.
What is your take on this discussion?