Not that big of a hit depending on the implementation, and a web services implementation can greatly enhance scaling. I'm currently working on .Net web services implementations that are intended to scale to 10-14,000 TPS. One major advantage from a performance side is load balancing your requests across multiple instances, so 1-N DNN instances running the GUI portion could reference 1-N business services. A standard installtion causes the business logic to execute on the server hosting the GUI, not always desirable in a large scale implementaion.

Also, from a security standpoint if they hack your DNN site they have the passwords for your web services, but they don't encessarliy have the passwords for the databases that are behind them.

True, but their opportunities for mischeif will be limited to what's provided by the web service API's, and a well designed site wouldn't allow you to call those API's from the outside world. They would have to get code onto the DNN site to call those services. Possible if they get host access, but that means learning the WS APIs (something somewhat difficult without access to the code), developing a module, installing it, etc.

Obviosuly the web services solution is not for typical isntallations. There's little point in doing it for anything but the largest scale inplementations.

if you segregate the business data from the data running the GUI then the hacker could destroy the site, but not still not gain access to the business data. Access to the business data would be through only the defined service interfaces, no SQL possible from the GUI instance, even with host access.

They would have to return individual objects or collections of them, and mimic parameterized calls for updates and deletes. Since they may or may not even be generating SQL on the other side it'll be a lot more difficult to inject. For example our data access layer uses a web services manager that routes based on the authentication token. They may retrieve data from different databases depending on who they are. In any case hacking is more difficult because the isolation means someone ca'tn just log in as host and start running commands to look at what's there in the SQL window. They would have to determine what to call, and what the WSDL was.

That is an excellent idea.

 Are there any examples of how to do this?