The debate over REST vs. SOAP is really an age-old one. However it fired up again over a recent remark by Tim Bray, the XML guru in Sun technologies. Tim said in an interview at OSCON that,
As before, proponents on both sides spoke up to support their favorite styles. The debate has formed a long thread with over 150 replies on the Service-Oriented-Architecture Yahoo! Group, where Nick Gall gives one example of a large enterprise that moved away from SOAP:
The SOAP stack is generally regarded as an embarrassing failure these days ... REST does what [the SOAP stack] was trying to do in a much more viable, elegant, cheap, affordable way except that we've got no tooling around it yet.
Wal-Mart replaced its supply chain's VAN EDI infrastructure several years ago with EDIINT AS2 and is still happily using it to this day. AS2 is basically POX, with its own approach to idempotence for reliable message delivery.
I've been predicting SOAP would never see widespread use outside the firewall.
And when talking about what can be considered a successful example of using SOAP, Nick points out:
What I am really seeking are large enterprises that are truly leveraging the power of SOAP in ways that provide convincing evidence that SOAP "works for the job" in ways that other approaches would struggle. IME very few enterprises really need SOAP for what they are doing -- it was put either put in by consultants as a checklist item, or the tool used SOAP by default. The majority of SOAP use appears to be simply driven by inertia, not any belief in its superiority in doing the job.
You have given us an example of a customer using REST for B2B - as an EDI replacement. EDI is document-oriented, as is REST, so this is not a big surprise.
We know of many examples of large enterprises using REST that designed and created their data centers based on the Web, since they are basically Web oriented businesses. But I do not know of any examples where an enterprise whose data center predates the Web has adopted REST.
...
Me - I am a "right tool for the job" person, and I cannot believe that everyone will be better off using REST any more than I can believe everyone will be better off using Web services
Besides, Steve Jones doesn't agree with the view that SOAP is not adopted due to its complexity, he notes:
I never get the complexity thing, REST isn't simple its got some very "nice" bits in it that are far from simple and its got some tough challenges (security for example) that aren't addressed. SOAP isn't complex, the complaint people make about it is that it is too easy and hides the network. So not doing SOAP might have many reasons but it can't be both complex and too easy.Obviously this will be another endless debate over the REST vs. WS-*/SOAP issue, although David Chappell claimed "the REST vs. WS-* war is over". People have been expecting a real end to the long debate, but it will not be a "crushing victory for one side". As John Evdemon says,
One size doesn't fit all - use what best meets your customer's needs and be done with it.Gervas Douglas proposes a stack comparison between REST and SOAP to remove the significant mismatch of comprehension between the two camps. But there are different opinions regarding the selection of the reference model. While Steve suggests taking the SOA RM as the basis for a SOA stack and then mapping REST/SOAP onto a single model, Mark recommends using OSI. Anyway, Gervas has created a wiki entry for this stack comparison purpose, let us wait and see what will be happening there.
