HTTP is a reliable protocol, since it's based on TCP, which adds reliability to the protocol stack.
In what way do you perceive HTTP to be unreliable?
REST is a different method of programming/using WS; it adds nothing in terms of reliability (and actually diminishes the security that the SOAP stack has by way of WS-Security - which may be just fine for some or many applications).
Tom: If you need to make sure that an HTTP request goes through, you can usually just keep sending the request until you get a response. The HTTP standard puts restrictions on what the server can do in response to a GET, PUT, or DELETE request, so nothing bad will happen if the request is processed multiple times. (Assuming the service is well designed.)
However, those restrictions don't hold for POST requests. There are other ways of making POST requests reliable (like Post Once Exactly, which we cover in RWS), or replacing POST with PUT.
Live a little! The night is young! And we have umbrellas in our drinks! This umbrella has a tiny ad: