I am trying to make an automated test/load test on a Silverlight based webapp. AFAIK, Silverlight uses WCF to make web service calls to the server. Instead of trying to make those web service calls, currently I am considering capturing all http traffic between the browser and the server (using a proxy) and replaying that using a script (or perhaps a load test application) to simulate the users actions.
I am not an expert in this, but IINM all web service calls are ultimately sent as ordinary http requests and responses. So capturing those requests and replaying them should simulate the actions of a user interacting with the application as far as the server is concerned.
So my question is, is there anything here that I am missing? Is there any case not covered in this approach that would force me to use WCF calls from the code instead of plain http requests?
(This is a follow on to my earlier question)
Ultimately, what is put on the wire by your client is what is used by the server. Even if your client uses all sorts of wcf client behaviors and other things that change the messages in the pipeline before reaching the wire, that will be included. The only thing I can think of you could miss is transport-level stuff that gets negotiated before the messages are sent, like security negotiation and credentials. I’d give Fiddler2 a try, you can export recorded sessions to visual studio web tests.