A browser that doesn't show the exact URL of a page is like a car with a built-in GPS that doesn't show the exact location where it's at. After all, who cares about street addresses? The address is occuplied by a Starbucks and we're in Mountain View, so let's just show a Starbucks logo surrounded by a shape that vaguely looks like an outline of Mountain View. You want to go someplace else? We'll show you your destination and the series of turns you need to make to get there, but we won't show you anything else on the way.
If you think a preschooler doesn't need to know how URLs work, you are vastly underestimating the curiosity of a typical preschooler. If you show him a bar with a bunch of letters in it, he'll start typing random letters into that box to see what happens. Likewise, if you show him a detailed map of the town, he will want to explore parts that he's never been to so far. Tinkering and exploration are the foundation of every science, including computer science. Therefore, I don't think it's a good idea to discourage tinkering and exploration, whether in a car or in a browser, unless the benefits greatly outweigh the long-term costs.
With cars, the benefits are probably quite large, since complexity is exactly what makes modern cars so safe and efficient. With browsers, I'm not even sure what the benefits are supposed to be, other than the obvious financial benefit to Google. People who don't want to tinker with the URL bar will just ignore it most of the time. Also, we're talking about the desktop browser here. There are plenty of pixels to waste.
Latitude/longitude would be more like IP addresses and HTTP headers. They require some technical knowledge to use and understand, but they're still quite human-readable unlike raw GPS signals or ethernet frames.
(As an aside, most GPS units also display the altitude. I live in a mountainous region, so I often make use of this figure.)
I'd like that to be true, but I think we lost that battle a long time ago. Google results aren't a readable URL; nor are products on Amazon or Ebay or anywhere else I can think of. Newspaper-type URLs are often "fake human-readable"; the URL is something like http://somepaper.com/12345-Local-Man-Found , but in fact http://somepaper.com/12345-Local-Man-Still-Missing will give you exactly the same story. Even HN stories aren't human-readable, just an opaque id number.
But I don't think "fake human-readable" URLs break the analogy with physical addresses. There are many different ways of writing the same address:
987 Some Avenue West, Unit 123, Brooklyn, New York, NY 12345-6789
Unit 123, 987 W. Some Ave., New York 12345
123-987 Some Av W, NYC, NY
And the numbers aren't just opaque identifiers (except for the zip), at least if you're walking down the street: you know that 28 is next to 26, opposite-ish 27, and halfway to 56. There's nothing that corresponds to walking along the street on a website.