Question

Also, how do connection strings fit in?

Presumably, it is ok for the dal to read these values from the config, or should these be passed down from the top?

Answer 1

I tend to say yes to this.

There are several reasons for encapsulating your custom configuration in a class and then injecting it, including:

• Testability1 – your config can be mocked out easily. Any classes under test that require config, do not require you to play around with App.Configs just to get the tests working – your are nicely decoupled from Microsoft’s ConfigurationManager
• Testability2 – the configuration itself becomes testable. If you are doing any “playing around” with values (e.g. turning a comma seperated list of values into an array) in your config class, you can test it’s behaviour simply and in isolation.
• Swapability – you can easily swap out your config provider to be config in code, or config from a database because the actual configuration is encapsulated.

So yes 🙂