After 21 months of public development, CoreWCF has finally reached its first GA release. It’s been a long path to get here, and thee is still have a long way to go. CoreWCF started off as a prototype internal to Microsoft which was started after a meeting with Scott Hunter. During that meeting Scott said something which has shaped the entire development of CoreWCF. He said that we only want one service host (the concept, not the WCF ServiceHost class) in .NET Core. ASP.NET Core had already been released, which meant there wasn’t a place for ServiceHost (the WCF class). This made a lot of sense, so I set about building a WCF prototype on top of ASP.NET Core. Eventually a decision was made to release the prototype as a community project under the name CoreWCF. The prototype focused on the developer experience so a lot of the implementation needed some love and attention, and 21 months later, we’re releasing our first GA release. Along the way, the CoreWCF project caught the attention of the Amazon AWS team and they have been contributing significantly to help make this project a success. It is thanks to the contributions from Amazon AWS that the initial release supports WS-* protocols (WS-Security, WS-SecureConversation), and some token authentication credential types with the TransportWithMessageCredentials security mode. This is a significant step towards supporting WS-Federation to enable moving enterprise WCF services to CoreWCF hosted on a cloud platform.
So what changes happened that took 21 months you might ask? Some fundamental architecture changes to make this project sustainable and cross platform. There have been 2 major themes to the changes. Removing Asynchronous Programming Model (APM) api’s and code, and removing direct native system calls and IO code.
The APM programming pattern is incredibly fast and squeezes out every last bit of performance, but at the cost of code maintainability and debuggability. WCF uses APM to its limit and the codebase can be hard to work with. Porting the code over as-is would require a large commitment from anyone wanting to contribute anything non-trivial. This also would make it hard for the community to be self-supporting. CoreWCF now uses async/await Task based asynchronous programming throughout. Computers are so much faster today than 10 years ago. For a community owned and supported project maintainability is a high priority than speed, within reason. CoreWCF also switched to a request push pipeline model adopting the ASP.NET Core middleware pattern. This made the code simpler, but required a lot of refactoring and rewriting of pieces of WCF.
The second big change is removing platform specific code and removing IO code. CoreWCF doesn’t even know what a Socket is and yet supports NetTcp. ASP.NET Core handles all of that for CoreWCF. It just reads and writes to pipes or streams. This means CoreWCF developer time isn’t spent having to handle writing code for some obscure scenario on a specific platform, it can (mostly) let the ASP.NET Core team worry about that.
All this is to say that supportability and maintainability is a high priority for CoreWCF.
A lot has changed, but a lot is also the same. Where changes have been needed, a concerted effort has been made to make things familiar. To give an example, let’s look at some code. When creating a self hosted service (console or WPF/WinForms as opposed to IIS hosted) in WCF, you would create a ServiceHost, add a ServiceEndpoint, and open the service:
With CoreWCF, there’s a little more code required due to the configuration model of ASP.NET Core, but the CoreWCF pieces are minimal. Those familiar with ASP.NET Core know that when configuring your host you need to specify a startup class which contains two methods, ConfigureServices and Configure. The ConfigureServices method is where anything needed in the dependency injection (DI) system is added, and any extension methods are called to add support for various frameworks such as MVC. At it’s simplest, a single method is called to add WCF support.
In the Configure method, things will look a little more familiar. The extension method UseServiceModel(this IApplicationBuilder app) is called with a delegate which configures CoreWCF.
CoreWCF still supports the System.ServiceModel namespaced contract attributes such as ServiceContract and OperationContract if the WCF Core client nuget package System.ServiceModel.Primitives is referenced. This allows sharing contract assemblies between the client and the server. There is also CoreWCF namespaced versions of these attributes too so a dependency on the WCF Core client nuget package isn’t needed.
There are multiple resources to find out more and to ask any questions about CoreWCF:
- Sample usages of CoreWCF alongside their WCF conterparts are available here.
- Discussions are enabled on the CoreWCF repo here.
- Release roadmap can be found here
- Feature prioritizing issue where you can vote for what you need is here
- Monthly community sync-up meeting details can be found here
- CoreWCF gitter channel can be found here