Re: Compatibility
Who on earth thought that 128-bit addresses was a good idea?
The idea of 128-bit addresses is you can afford to waste some of them in order to simplify routing. IPv6 addressing breaks down on some very specific barriers. There's no "wait, what's the broadcast address for a /27, again?" mental math.
The biggest thing to remember is that the smallest subnet is a /64. The last 64 bits are yours to do what you want with, so you can make them something memorable if you like. So that's a maximum of 16 arbitrary hex digits you might have to remember. But! You also get to omit leading zeroes, and collapse 16-bit blocks that are all 0. So it's not even that bad.
To give a real-world example, Comcast's DNS servers are 2001:558:feed::1 and 2001:558:feed::2. Those don't strike me as harder to memorize than most arbitrary IPv4 addresses. (OK, they're harder than 8.8.8.8, but that's something of a special case.)