问题
Most of the WCF examples out there show you how to configure WCF client and server. Now, what happens if you differ the configuration slightly between them? I mean, who has the final word?
Let's take this client configuration for example:
<configuration>
<system.serviceModel>
<bindings>
<wsHttpBinding>
<binding name="WSHttpBinding_ISampleService" closeTimeout="00:01:00"
openTimeout="00:01:00" receiveTimeout="00:01:00" sendTimeout="00:01:00"
bypassProxyOnLocal="false" transactionFlow="false" hostNameComparisonMode="StrongWildcard"
maxBufferPoolSize="524288" maxReceivedMessageSize="65536"
messageEncoding="Text" textEncoding="utf-8" useDefaultWebProxy="true"
allowCookies="false">
<readerQuotas maxDepth="32" maxStringContentLength="8192" maxArrayLength="16384"
maxBytesPerRead="4096" maxNameTableCharCount="16384" />
<reliableSession ordered="true" inactivityTimeout="00:10:00"
enabled="false" />
<security mode="Message">
<transport clientCredentialType="None" proxyCredentialType="None"
realm="" />
<message clientCredentialType="Windows" negotiateServiceCredential="true"
algorithmSuite="Default" establishSecurityContext="true" />
</security>
</binding>
</wsHttpBinding>
</bindings>
<client>
<endpoint address="http://localhost:8080/SampleService" binding="wsHttpBinding"
bindingConfiguration="WSHttpBinding_ISampleService" contract="ISampleService"
name="WSHttpBinding_ISampleService">
</endpoint>
</client>
</system.serviceModel>
Usually the server side will have the exact same binding configured on its exposed server.
But what happens now if on the server side is defined with openTimeout = 00:00:30. What will be the timeout? Who wins? I do the same question for all other parameters.
The whole thing seems a big mess, how can you tell, for each element of the configuration (binding, client, service, behavior, etc.) and all their details, which parameters is required and in which side (client or server)?
It seems you could define the entire binding with all timeout parameters on the server side and on the client side simply put the minimum required configuration so all parameters from the server are accepted. But now, what are the minimum required parameters on the client considering the server has a more in depth configuration?
What are the best practice when configuring client and server using WCF regarding parameters for each element of the configuration: bindings, services, client/endpoint and behavior?
When conflicting parameters are defined between client and server how WCF handles it?
回答1:
In order to address your request in your last comment to my previous answer, I tried to come up with my approach to how I would create (and modify) server- and client-side config's for any given service. This is based on both theory I read (books, blogs), things I've learned in Juval Lowy's WCF Master Class, and quite a bit of practical experience with several large service implementation projects - this isn't available in one single place, on the web or in a book.... so here it goes:
I would start basically from scratch. Think about your service first:
- what address does your service live at?
- what binding(s) do you want to support?
Simplest scenario: single service, single endpoint, basicHttpBinding, all defaults
Service config:
<system.serviceModel>
<services>
<service name="YourNamespace.YourServiceClass">
<endpoint name="Default"
address="http://YourServer/SomeVirtualDirectory/YourService.svc"
binding="basicHttpBinding"
contract="YourNamespace.IYourServiceContract" />
</service>
</services>
</system.serviceModel>
Corresponding client config:
<system.serviceModel>
<client name="Default">
<endpoint name="Default"
address="http://YourServer/SomeVirtualDirectory/YourService.svc"
binding="basicHttpBinding"
contract="YourClientProxyNamespace.IYourServiceContract" />
</client>
</system.serviceModel>
Then only ever change something if you really must! And most of all: NEVER EVER let Visual Studio (Add Service Reference) or svcutil.exe screw up your config! Protect it like the apple of your eye!
Then: if e.g. your data transfer takes more time than the default timeout of 1 minute allows, change that one single setting on both the service side and the client side. Do this by defining a custom binding configuration and referencing that from your endpoints - but change only that - not more! Leave everything else as is, with default values. Don't ever change anything unless you absolutely must (and know what you're doing, and why you're doing it).
Mind you: the sendTimeout on the client (time allowed until the whole message has been sent) will correspond to the receiveTimeout on the server - the time allowed for the whole message to come in (see this excellent blog post and this MSDN forum thread for more information)
Service config:
<system.serviceModel>
<bindings>
<basicHttpBinding>
<binding name="ExtendedTimeout"
receiveTimeout="00:05:00" />
</basicHttpBinding>
</bindings>
<services>
<service name="YourNamespace.YourServiceClass">
<endpoint name="Default"
address="http://YourServer/SomeVirtualDirectory/YourService.svc"
binding="basicHttpBinding"
bindingConfiguration="ExtendedTimeout"
contract="YourNamespace.IYourServiceContract" />
</service>
</services>
</system.serviceModel>
Corresponding client config:
<system.serviceModel>
<bindings>
<basicHttpBinding>
<binding name="ExtendedTimeout"
sendTimeout="00:05:00" />
</basicHttpBinding>
</bindings>
<client name="Default">
<endpoint name="Default"
address="http://YourServer/SomeVirtualDirectory/YourService.svc"
binding="basicHttpBinding"
bindingConfiguration="ExtendedTimeout"
contract="YourClientProxyNamespace.IYourServiceContract" />
</client>
</system.serviceModel>
As you need other changes, like multiple endpoints on the service side, or local settings like bypassProxyOnLocal - adapt your config, do it carefully, step by step, manually, and consider your config an extremely essential part of your whole service - take care of it, put it in version control etc.
回答2:
If a given timeout is about the same thing in the end, on the client and the server, and the two values don't match, the shorter timeout "wins" - no matter whether it's defined on the server or the client.
Most other things like addresses, bindings etc. must match - otherwise, no communication will be possible...
The benefit of the server setup is that you can define multiple endpoints for a single service, with different options and bindings - and the client can then "pick and choose" which endpoint to connect to.
But once one service endpoint has been chosen for your connection, the settings like bindings, protocols, security, reliability and so forth must match exactly.
Also: the default config that the Visual Studio Add Service Reference or the svcutil.exe command line tools generate are both catastrophically bad - they contain way too many settings that all reflect the default values, which make it really hard to know what's really needed and what is not.
In that respect, I'd recommend watching the DotNet Rocks TV Show #122: Miguel Castro on Extreme WCF in which Miguel nicely shows just how easy it really is to create those configs manually, and then you completely understand what the heck you're doing! Highly recommended!
来源：https://stackoverflow.com/questions/4879310/when-setting-up-a-wcf-client-and-server-how-synchronized-does-the-config-files