The protocol itself is uncensorable, right? Any relay can join and offer its notes to the network.

The flip side is that filters of all sorts will be added within individual relays and clients - could be for spam, for content that is offensive for some, taste, etc. Different audiences will each want different filters. (Some want 0 filters).

This is a radically new paradigm, and figuring out sensible ways to make it work will be #nostr's biggest challenge. I would argue that transparency is key: you should know what is being filtered out, which criteria are in play