As much as I like WCF (and as much as I made money with it in the past years) – I think it is now time to re-think some approaches and some architectural ideas. What is ahead of a lot of us is what I call ‘light-weight architectures for reach’.
Just to give you some scope: I am one of the original Digerati board members of Indigo (WCF). Some people say I know WCF inside-out - I have given several dozens of introductions to WCF in seminars, conference and user group sessions world-wide and did numberless consulting gigs in Europe with WCF (and no, I did not really find time to write a book – besides a chapter here).
OK, so here goes...
There is currently a shift - or rather a new focus - in distributed application architecture. And no, I am not going to call it REST and I am not saying it is the one and only true thing from now on. But we have new drivers like application mash-ups, mobile devices and cloud computing which force us to move away from the good old feature-rich (and somehow heavy-weight) SOAP-based approaches. This development and this need is non-discussable and undeniable. One possible solution to these new needs is the use of light-weight Web APIs (which can be evolved into a REST-ful API or be used 'just' in an RPC-ish way).
So, which technology framework should we as .NET developers choose to build Web APIs? One choice is to leverage WCF's WebHttp model. After working with it for some years I quickly noticed a number of shortcomings (testability/mockability, serializers, ...) and wished that Redmond would come up with something better. *Of course* it had to be WCF, it had to be part of the overall "let's abstract away communication 'details'". And so it happened that they announced WCF Web API (and I was lucky enough to be part of the advisory board, again). All was fine in my world. It seemed.
Fine until I again realized that it has shortcomings when building some more flexible and advanced Web APIs. But this time it was actually the fact that the limitations were in WCF itself. The 'SAOP message' model of WCF just does not fit at all into this world. Everything the team tried to accomplish and to provide features for a web-based paradigm was just trying to put the round into the rectangle (and vice versa). For a web-based base framework maybe WCF is not the right choice.
And it turned out that there is a very good framework that can deal with the web and HTTP in a fantastic way. So, the decision to build a .NET web API framework on ASP.NET (and ASP.NET MVC, to be more precise) was somehow natural.
I personally had one very strong wish - something I have learned in the past years is invaluable for a lot of our customers... self-hosting! So, the advisors pushed hard to get self-hosting, and here it is: we can perfectly build our own host processes and host the exact same Web API implementation either there or in IIS/ASP.NET.
Today, I am quite happy with what happened. The only issue I still have is the name: I would definitely have not called it ASP.NET Web API, but rather .NET Web API. If you are working in my geographical area then you may understand why.
Anyway... talking a bit about application architecture (.NET server-side):
I would build a WCF service and an ASP.NET Web API (at least with the RPC-ish approach) as a facade only - the actual logic (business logic or data/resource access) is hidden beyond that facade. With this simple pattern it is no problem to have both WCF SOAP services and ASP.NET Web APIs sitting right next to each other happily ever after. I am doing it all the time. And you are then actually in the power to leverage the full power of each framework - get the most out of SOAP/XSD/WSDL/WS-* and the best out of web APIs/REST/whatever.
Fact is, I am using Web APIs quite a lot these days as we are building cloud-based systems in an increasing number and also cross-device mobile apps.
My 2 services cents.