It helps if you think about it like this:

There is the inside of your program, and there is the outside world, and there is (or at least, should be) one clean and definite interface between the two. In the inside of your program you use strings that consist of (potentially wide) characters. The outside world deals with a stream of octets. When you read something from the outside world into your program, you convert (decode) from octets to a string of characters, and when you emit something from your program to the outside world, you convert (encode) the characters into a stream of octets.

The precise mechanics of those encode/decode steps depend on the application and the kind of protocol used. There isn't (and cannot be) a single answer whether you should or should not use Encode::encode("UTF-8", ...).

In your HTTP example, the need to encode depends on what the other side requires. Again, we can't tell in general. Presumably, it's documented by the service you're trying to POST to. The encoding you use should match the contents specified in the Content-Type header: if it's text/html; charset=UTF-8, then you must encode as UTF-8, if it's e.g. image/png, you must not.