[You should read Part 1 of this little series before you proceed reading this one.]

In this 2^nd part I am extending the simple messaging example of Part 1 by adding some explicit WS-Addressing trickery. Addressing is so fundamental that its properties are baked right into the Headers collection of the Indigo Message. Even though there are (and I will eventually show) much easier ways to do request/reply management that hide most of what I am doing here very conveniently under the covers, I’ll give you an example of how you can send messages to a service and then instruct the service to explicitly reply back to an endpoint you provide. To make it a little more fun, I am setting up two alternate reply endpoints and have the service come back to them in turns. The Program host class is identical to the one the previous example, so I’ll show only client and service code along with the config.

The server-side code below grew a little bit as you can see. Now, there is a IGenericReplyChannel that is the contract for the replies. It looks suspiciously like the client-side’s IGenericMessageChannel and it is indeed a copy/paste clone of it. I just didn’t want to share code between client and server side. The Receive method has changed insofar as that it no longer prints the message to the console, but now creates a reply and sends the reply to the endpoint that the client indicates through the (WS-Adressing-) ReplyTo header of the incoming message.

To do this, the service constructs a ChannelFactory< IGenericReplyChannel>, using the endpoint address indicated in the incoming message’s ReplyTo header and getting the binding information from the “replyChannel” client setting in the config file shown further down. (Note that this is a bit simplistic, because it assumes that the ReplyTo EPR uses a compatible binding. There is a brilliant way to fix this, but … later). Then, the message body of the incoming message is read into an XmlDocument and if this was a real application, it would likely do something here. For now, we just leave the content as it is and punt it back out.

To construct the reply message, I don’t use the CreateReplyMessage() method provided on the Message class, simply because it doesn’t have an appropriate overload to deal with an XmlReader in the same way as Message.CreateMessage() does. I am sure that’s a minor oversight that’s just a problem with my particular Indigo build. Creating a reply is quite simple, though. All I need to do is to copy the incoming message’s MessageID value into the RelatesTo.Reply property of the outgoing message. For simplicity, I don’t check whether that header is present and set, which I really should do, because there is no actual contract or policy in place (for now). Once I have the reply constructed, just copying the incoming body into it, I send it out through a channel (“proxy”) constructed by channel factory.

Having a reply-enabled server-side, we can now get to the juicy part: the client. Since we now need to listen for replies, the client has to expose a reply-endpoint and therefore also act as a server. (That is the reason why “endpoint” is preferred in service-land rather than the “client”/”server” nomenclature). Therefore, I define a IGenericReplyEndpoint contract (no surprises there) and implement that in GenericReplyEndpoint. To make the example a bit more fun, the constructor of that service class takes two arguments: The client argument refers to an instance of the Client application class and epName gives the service instance (!) a name. The client reference is used to let the client application know how many messages were already received so that it can shut down, once the expected replies for all sent messages have come back. The notification about received messages is done inside the ReceiveReply method,  which otherwise just writes the message body to the console.

Unlike the previous example, this service implementation isn’t used directly. Instead, I derive two subclasses from it: ReplyEndpointA and ReplyEndpointB. These two classes each implement a constructor that passes “A” and “B”, respectively, for the epName argument to the base-class and pass-through the client argument. In case you wonder how the ServiceHost could possibly construct instances of these service classes, not knowing the appropriate parameters to pass to them: Instances of these two classes are pre-constructed and fed into the service host as singletons as you will see below.

The Client application class is a bit more intricate than the previous version, but there is no rocket science there. I have a counter for the number of messages received and a ManualResetEvent that is getting signaled whenever the number of received messages matches (or exceeds) the number of sent messages. That happens in the MessageReceived method, which is called by the service singletons. The class also has a UniqueIDGenerator, which is an Indigo-supplied class that lets me generate values for the MessageID header that is required alongside using ReplyTo.

In the SendLoop method, I now create two service host instances that shall receive the replies to messages I send; one of type ServiceHost<ReplyEndpointA> and one of type ServiceHost<ReplyEndpointB>. Each of these hosts receives an instance of its service type as a construction argument. Doing so causes the service host to operate in singleton mode, meaning that it will not create new service instances out of and by itself, but rather use only the exact instance supplied here. In the actual send loop, I alternate (i%2==0) between those two service hosts and invoke SendMessage passing the channel factory (not the channel as in the previous example) and the chosen ServiceHost instance.

In SendMessage, I do a few simple things and only one not-so-obvious thing. A new message is constructed as the first step and loaded with an action and the body content. Then I grab the destination address from the channel factory, which sits in the channel factory’s Description.Endpoint.Address property and assign that to the message’s To header. The MessageID is set to a new unique identifier created using the messageIdGenerator. All that is pretty straightforward. Not immediately clear might be what I am doing with the ReplyTo header:

Once a service host is Open, it’s bound to set of endpoints and is actively listening on those endpoints using “endpoint listeners”. I am writing “set of endpoints”, because a service might have several. Each service can expose as many endpoints as it likes; each with a separate binding (transport/behavior/address) and each with a separate contract. There are puzzling special cases, of which you’ll see at least one in this series, where a service listens and properly responds to a contract type that is nowhere to be seen on the actual service implementation. The active endpoints sit on the EndpointListeners collection.

For simplicity (again, this is a bit naïve, but serves the purpose for the time being) and to obtain a ReplyTo address to pass to the service I am sending the message to, I reach into that collection and grab the first available endpoint listener’s address. What I should be doing here is to check whether that listener is indeed the one for the IGenericReplyEndpoint contract and whether I can find one with a binding that is mostly compatible with the one the outbound channel uses. The latter selection would be done to make sure that if I send out via “net.tcp” and I expose a “net.tcp” endpoint myself, I would preferably pass that endpoint instead of a possible “http” endpoint I might be listening on at the same time. Once ReplyTo is set, I send the message out.   

What’s left is the matching configuration for this. The mechanics of how the configuration maps to the classes and instances are largely the same as in the simple messaging example. A small difference is that the replyChannel client configuration has no target address attribute, because that one is always supplied via ReplyTo (refer to the GenericMessageEndpoint’s Receive method above to see how that is wired up). Oh, yes, and I switched it all to http transport in case you don’t notice. TCP would work just as well, but I felt like I needed a little change. The assumed assembly name for this sample is “SimpleAddressing”, of course.

The output of the sample is predictable, isn’t it? The replies come back in sequence, alternating between the two reply services “A” and “B”.

What I’ll start with is the most simplistic and “raw” way to use Indigo that’s practical for someone with a life. The extensibility model will let you reach even deeper down into the guts, but I don’t want to drag you down there too far. Also, I don’t really know all of the scary dragons lurking there is the dark. I also want to start with this example to dispel some people’s impression that Indigo is just another “square brackets RPC-ish thing”.

The snippet below is likely the simplest possible Indigo service.  I define an IGenericMessageEndpoint contract that has a single operation Receive, which expects an System.ServiceModel.Message as input. The Message class is the immediate representation of a message that the entire Indigo infrastructure uses internally and that can be surfaced to the application using a contract definition like this. The [OperationContract] attribute signals that the operation is “one way”, so it’s clear that we’re not sending any immediate responses and not even faults. The Action is set to “*”, which is a wildcard indicator specifying that all messages, irrespective of their Action URI will de dispatched here. That is, unless there would be another operation without a wildcard Action. In that case, all messages matching the concrete Action URI of that operation would be dispatched there and all other messages would flow into the wildcard operation.

The implementation of the contract in GenericMessageEndpoint just dumps the content of the message body onto the console by acquiring the XmlDictionaryReader of the message and writing the string’ized body content out.

The Server class constructs a ServiceHost<GenericMessageEndpoint> for the service implementation, which constructs endpoints from configuration settings, hosts these endpoints, and is responsible for creating instances of the service as messages arrive and need to be dispatched. As you can see, I do nothing more than constructing the host and Open it. The specifics of what transport is used and where the service is listening will be supplied in config, as you’ll see further down.

Below is the matching “raw” client. What we want to do here is to just construct a System.ServiceModel.Message, put some XML into it and throw it over the fence. To do that, I construct a client side contract IGenericMessageChannel (I am doing that to show that we’re really in “contract-free” raw messaging territory here) that has a Send operation, which “looks right” on the sender side vs. the receiver contract’s Receive, and also flags the operation as one-way and with a wildcard Action.

To setup a channel to the destination service, I can now (in SendLoop) construct a ChannelFactory<IGenericMessageChannel> over that contract and with the argument “clientChannel”, which is a reference into the configuration as I’ll show in a little bit. The channel factory is the client-side counterpart of the service host. It reads all information about the channel from the configuration, evaluates the bindings, binds to the right transports and behaviors, and also knows about the endpoint to talk to. Once I have a channel factory, I can Open it and have it give me a channel (or “proxy”) that I can talk through. In SendMessage I cook up a Message from an Action URI that I make up and an XmlReader instance layered over an XmlDocument that I keep around and send that out to the service.

The surrounding application for Client and Server (I run both in the same process for simplicity) is, of course, trivial. All I do is to construct and start the server, construct a client and call its send loop and then wait for the user to be amazed of the (server’s) console output and have him/her press ENTER to quit. If I were making this more elaborate, I could wait until all sent messages had arrived at the service side and shut down automatically, but this is supposed to be simple.

That’s as much code as we need to implement a one-way messaging client/server “system” that can throw XML snippets across a network transport.

To make it work, we need to configure this application and “deploy” it to a concrete environment. A simple configuration (assuming this is all compiled into “SimpleMessaging.exe” and hence the assembly name is “SimpleMessaging”) could look like the one shown below.

The <bindings> section contains one <customBinding> (means: I am not anything predefined), with a concrete configuration named “defaultBinding” that uses the tcpTransport. If I were setting up security or reliable messaging, would also be doing that here and add the respective config elements alongside the TCP transport binding element, but we will keep it simple for the time being.

The <client> section defines, for the configurationName=”clientChannel” (look above in the client snippet how that maps to the ChannelFactory< IGenericMessageChannel > constructor call), which binding should be used. The example links up to the customBinding type and within that type to the “defaultBinding” config. Furthermore, the section defines to which contract type ([ServiceContract]-labeled interface or class) the endpoint is bound and, lastly and most importantly, the address at which the endpoint is listening to client messages.

The <service> section defines the server side of the story. The association between the service host and the configuration is done via the serviceType attribute. When the ServiceHost<GenericMessageEndpoint> is contructed in the server snippet above, the service host locates the section for the respective service type it is hosting by looking at this attribute. The endpoint definition on the server side is very similar to the client side, which should not be very surprising. It also refers to a binding using bindingType/bindingConfiguration, defines the address at which the service will be listening, and indicates which contract type applies for the endpoint.

Running it all yields the following output, spit out by the server side, and just as expected:

Very simple and versatile one-way messaging flowing free-form XML. Not at all RPC-ish.

The scariest search for which my blog is on Google rank #1 is power=work/time. I knew that Google loves me, but this starts to become pretty ridiculous

CNet reports about Bill Gates’ announcement that Windows Anti-Spyware is going to be free includes the following truly puzzling quote from the Check Point Software CTO:

"I am glad to see Gates is focusing on securing the desktop," said Gregor Freund, chief technology officer of Check Point Software, which develops desktop security software. "However, there are some serious downsides to Microsoft's approach. Just by entering the security market, Microsoft could stall innovation by freezing any kind of spending of venture capital on Windows security which in the long run, will lead to less security, not more."

Is it just me or do you also consider the term “venture capital” as being a little out of place in this context?

Jim Johnson, who works on transaction technology at Microsoft’s Distributed Systems Group (aka „Indigo team“), shares an important insight on ACID’s “D” and that “Durable” doesn’t always mean “to disk”, but is rather relative to the resource lifetime. A transaction outcome can be “durably” stored in memory, if that’s what the characteristic of the resource underlying the resource manager (=transaction participant) is. That’s a very important perspective to have. Once you get away from the “transactions are heavy duty” thinking (because “D” seems to imply “disk” to many people), transactions, especially with “lightweight transaction managers” such as the one found in the .NET Framework 2.0’s System.Transactions.Ltm namespace (or the one I published a while back), suddenly become very attractive to coordinate and resolve fault conditions between components within a process.

Aaron Skonnard says I am clearly wrong with my demand that one shouldn’t have to look at WSDL/XSD/Policy. Well, at this point in time the tooling makes it indeed difficult to ignore angle brackets. But that’s not a reason to give up. I also find the “it all has to start at the angle bracket” stance overly idealistic.

I can type up an XML Schema in notepad, I can even type up a WSDL in notepad. As much as one would like to have it different, both “skills” are not so common amongst the developer population. I would think that for the majority of ASP.NET Web Services in production today, their developers completely ignored the XSD/WSDL details. But even if that were different: The rubber hits the road when we talk about policy. Can you type up a complete and consistent set of policy assertions for integrity and confidentiality and authentication using Kerberos and vX509 tokens without looking at the spec or a cheat sheet? How about combining that with assertions for WS-AT and WS-RM? As long as we keep the story reduced to XSD and WSDL, dealing with angle brackets might something that someone could reasonably expect from a mortal programmer who has a life. One we take policy into the picture, we better start asking for tools that hide all those details. The interoperability problems of getting secure, reliable and transacted web service work together are far harder than just getting services to talk. That’s part of the contract story, too. Yet, I cannot imagine that anybody would seriously demand that we all sit down and explicitly write these endless sequences of policy assertions and then feed our tools with them. At least I don’t want to do that, but that may just be me getting too old for this stuff.

Bruce Williams illustrates how to turn my very simple “Hello World” Indigo sample into a queued service by changing the transport binding from HTTP to MSMQ (I think that’s radically cool). Now, the next step is to illustrate a Duplex conversation to get the response back to the caller. If Bruce or someone else isn’t going to beat me to it, I’ll show that once I get home from Warsaw tomorrow night. [Ah, by the way: Bruce! No need to “Mr.” me ]

Tim Ewald responds (along with a few others) to my previous post about WSDL and states: ”Remember that WSDL/XSD/Policy is the contract, period. Any other view of your contract is just an illusion.”

WSDL and XSD and Policy are interoperable metadata exchange formats. That’s just about it. The metadata that’s contained in artifacts compliant with these standards can be expressed in a multitude of different ways. I do care about “my tool” (whatever that is) to do the right thing mapping from and to these metadata standards whenever required and I do care about “my tool” guiding me to stay within the limits of what these metadata formats can express.

But WSDL/XSD/Policy isn’t the contract. If you do ASMX, you can create server and client without you or any of the tools ever looking at or generating WSDL. And it works. If you use Indigo, you can do the same and, in fact, for generating any XML-based metadata from within an Indigo service, it’s even required to explicitly add the respective service behavior at present. The required metadata to make services work comes in many shapes or forms and is, for a given tool, typically richer than what you will find in the related WSDL/XSD/Policy, because not all that metadata is related to the wire format itself.

If I need to tell someone who is not using my tool of choice how to talk to my service, I have my tool generate the respective metadata exchange documents and I want to be able to trust my tool that they’re “right”.

What I am stating here is simply my demand and expectation for the degree of “automatic interoperability” that I expect from the tools. I can read WSDL/XSD/Policy; out there, most people absolutely don’t seem to care about these details and I tend to agree with them that making this stuff work is someone else’s problem.

I don’t need to be able to read and write PDF to use PDF. I use PDF if I know that someone will open my document who is not using Microsoft Word. Still, that PDF doc isn’t the document. My Word source document is the document I edit and revise. The PDF is just one of several possible representations of its contents.

I wish I was at VSLive! in San Francisco to hang out with all of my friends. Instead (and that isn’t too bad, either), I am sitting in my hotel room at the Warsaw Marriott watching the sun rise over the Polish capital. Today and the next two days, my partner Achim Oellers and myself will be teaching a class on service orientation principles, explaining fundamental ideas, patterns, techniques and will go through a lot of concrete implementation guidance for today’s Microsoft MSMQ/WSE/ASMX/ES stack so that our customers can start writing services today. The fundamental principles about data contracts, message contracts and service contracts that we teach will carry forward to Indigo – along with a lot of the implementation techniques (and the resulting source code) that we will suggest. Of course, that has been a bit of a hidden agenda in past workshops, because I couldn’t openly speak about anything that happened to Indigo past PDC03, but now that the Indigo day at VSLive! is over, I can. That makes it even more fun.

XML is ugly and angle brackets are for plumbers. Unless you have a good reason to do so, you shouldn’t have to look at WSDL. Sharing this C# snippet here

[ServiceContract]
interface IHello
{
      [OperationContract]
      string SayHello(string name);
}
is a perfectly reasonable way to share contract between server and client, if you’ll be sticking to Indigo. A service can expose all the WS-MetadataExchange and XSD and WSDL you like so that other Web Service clients can bind to your service, but as long as you stay on the System.ServiceModel level and focus on writing a distributed systems solution instead of writing something that “does XML”, you won’t have to worry about all the goo that goes on in the basement. Staring at WSDL is about as interesting as looking at the output of “midl /Oicf”.

using System;
using System.ServiceModel;

namespace IndiHello
{
      [ServiceContract]
      public class Hello
      {
            [OperationContract]
            public string SayHello(string name)
            {
                  return "Hello " + name;
            }
      }

      class Program
      {
            static void Main(string[] args)
            {
                  ServiceHost<Hello> host = new ServiceHost<Hello>(new Uri("http://localhost/hello"));
                  host.AddEndpoint(typeof(Hello), new BasicProfileHttpBinding(), "ep");
                  host.Open();
                  Console.WriteLine("Press ENTER to quit");
                  Console.ReadLine();
                  host.Close();
            }
      }
}

I am told that I can talk, so I do   Here’s a simple Indigo server. If you looked at the PDC 2003 Indigo bits, you will notice that the programming model changed quite a bit. I think that in fact, every single element of the programming model changed since then. And all for the better. The programming model is so intuitive by now that I am (almost) tempted to say “Alright, understood, next technology, please”.

So up there you have a class with an implicit service contract. An explicit service contract would be a standalone interface (that’s the proper way to do it, but I wanted to keep the first sample simple) with a [ServiceContract] attribute. Here, [ServiceContract] sits right on the class. Note that the class doesn’t derive from any special base class. Each method that you want to expose as an endpoint operation is labeled with [OperationContract]. These and a set of other attributes (along with a bunch of options you could set, but which I am not doing for the moment) control how the class contract is exposed to the outside world via Indigo.

In the Main method, you have a ServiceHost, which hosts the service (the class is parameterized with the implementation type) and which is initialized with the base-adress at which the service shall be hosted. The base address here is “http://localhost/hello” and with that maps into the namespace of http.sys at port 80. The endpoint can exist alongside any IIS-hosted websites, even though this particular app is hosted in its own little console-based app.

Into this host, I map the service contract with a BasicProfileHttpBinding() to the endpoint address “ep”, which means that messages to that particular service that flow through HTTP using the WS-I Basic Profile 1.0 shall be directed to the “http://localhost/hello/ep” endpoint. Once I have a binding in place (that could also be done in config), I Open() the service and the service listens. Once I am done listening, I Close() the service.

Isn’t too hard.

There you go:

Happy Stewardesses who like my Alienware notebook (seems to work just as well as driving a Lamborghini)

And ... chatting with Hanselman and having (economy class ... so much for Lamborghini) food