There’s been a splurge of REST vs. SOA (or WS-* vs ROA) discussion in the last couple of weeks that’s got my attention and made me think a little more deeply about it.
I really like the term ROA as opposed to REST as it makes a clearer comparison between 2 design styles rather than a design style (SOA) and what is generally an HTTP specific implementation (HTTP).
It was also good to see some recognition that both design and implementation methods can co-exist, that there’s not one xOA to best deal with all situations.
I agree that the Axis2 POX/HTTP stuff probably shouldn’t be labelled REST as there’s nothing there to ensure that your ‘Axis2 REST’ services actually are. Sanjiva thinks we just provide the gun and users pull the trigger. It seems to me more like we’re sticking a label of ‘water pistol’ on a gun and then blaming the user when someone gets hurt.
However, the suggestion that Apache doesn’t ‘get REST‘ is miles off. Apache Abdera looks really promising in the Atom space and the latest REST support in Apache CXF looks really, really easy to use.
So after a lot of investigation, where am I? Where I think everyone who isn’t trying to sell one of these should be. On the fence, evaluating the right technology to use for each individual scenario.
I’m not sure I really agree with everything he’s saying, but Adam Bosworth’s untitled talk here: (which is essentially about standards simplicity)
http://www.adambosworth.net/archives/000031.html
has a lot to say for me about the differences between SOAP, REST, etc.
I totally agree that KISS is good, that HTML is a success because of sloppy parsing, and that RSS is too… and that they’re great.
I think it comes back to using the right tool for the task.
When I think about a bank and credit card company exchanging information I don’t believe that these simpler solutions (e.g. RSS) have proven themselves yet. So how can we deal with that? We pump up the formalism.
Do I think that we can do all of these things more simply? Yes.
Do I think that if we tried today we (and our customers) could be as confident in the solution? No
Do I think we could have a WS-* 2.0 which takes a number of years of real-world experience and simplifies everything? In another couple of years, Yes
Do I think it will happen? Maybe