I think putting the high level API (which I really like the clean and simple design of, by the way) under something like
std.net and the low level curl specific API in etc.curl might be nice. I know what curl is but looking for std.curl to download a file isn't obvious for anyone who isn't aware of curl. This also leaves the option of keeping the high level API and swapping out the curl backend with something else if that was ever needed or wanted. That's not very likely to happen anytime soon but I think a good guideline might be to keep the phobos API focused on what, specifically, you want to do rather than what is doing what you want to do.