John Gilmore was talking about Usenet, where that was actually true; if some intermediate netnews server started deleting articles that contained bad words, people using other news servers wouldn't even notice unless they got their only newsfeed from the censoring site. (Later on, this became a problem, as it was impossible to censor spam as well.)
The internet is not like Usenet in that way. Its design is full of central points of failure: root nameservers, TLD nameservers, DNS registries, BGP, and AS and IP allocation, among others. Some newer network protocols like BitTorrent are indeed capable of routing around censorship and other forms of damage, but older ones like the Web are not.
I've written a bit about some thoughts on how to fix that, but I still haven't done much: http://lists.canonical.org/pipermail/kragen-tol/2006-Novembe...