I think the main difference is that webwormhole tries to use STUN/firewall hole punching to connect two nodes, while the original implementation will use a TCP-TCP relay as fallback.
I recently needed to transfer a 30 GiB file within a restricted network but the firewall blocked webwormhole, so I ended up with -williams instead.
Hopefully at some point all these project can converge into something that supports ALL the solutions :-)
* Native Client (in Go so you use pretty much everywhere)
* Uses PAKE for E2E security. Even if the server is hostile they can't get anything.