Therefore the overarching design goal of freenet is to make censorship as hard as technically possible. That’s the reason for - providing anonymity (else you could be threatened with repercussions - as seen in the case of the wikileaks informer from the army in the USA), - building it as free software network (else you could just shut down the central website, as people tried with wikileaks), - providing safe pseudonyms and caching of the content on all participating nodes (else people could censor by spamming or overloading nodes), - including good moderation tools (otherwise people could censor by just "flooding the zone with shit"), and even - the darknet mode and enhancements in usability (else freenet could be stopped by just prosecuting everyone who uses it, or it would reach too few people to be able to counter censorship in the clearnet).
One core difference between MaidSAFE is that Freenet provides communication. Content in Freenet stays accessible while people access it, without tracking accesses: When you upload something, you can imagine that a clock starts ticking. After a while (depending on file size) the content is no longer accessible. But if someone accesses it, their node "heals" that content, which resets that clock.
Also Freenet enables people to connect purely to their friends, so there’s no central point of failure or detection.
I once did a comparison to other projects for a funding proposal (that did not succeed, though) which should still give you an idea about the difference between Freenet and other systems: http://www.draketo.de/proj/freenet-funding/#sec-9
In case anyone doesn't believe me, here's the Freenet thesis which actually mentions both Gnutella and Napster in references. Hard to claim precedence when that's the case.
That paper isn't the original Freenet paper, it came later. This is the original: https://www.semanticscholar.org/paper/A-Distributed-Decentra...
Including decentralized download via HTTP using just three added headers and Tiger Tree Hash Exchange: https://rfc-gnutella.sourceforge.net/src/Partial_File_Sharin...
Scalability was limited to global search in around 5 million nodes (with the network spanning 40 million people at its peak), though, due to restrictions in the number of TCP connections Windows (private edition) supported.
https://www.youtube.com/watch?v=JWrRqUkJpMQ
The interview -- like most of my interviews -- goes into depth about the entire space, and also touches on Locutus, his new project.
PS: I like to interview people on topics related to what we build at Qbix and Intercoin.
Freedom of Speech and Capitalism (with Noam Chomsky) https://www.youtube.com/watch?v=gv5mI6ClPGc
Identity and Civic Engagement (with David Boulet) https://www.youtube.com/watch?v=GRzVj2W9WGM
Free Cities (Patri Friedman, grandson of Milton Friedman) https://www.youtube.com/watch?v=Lgil1M9tAXU
Community Economics (Thomas Greco) https://www.youtube.com/watch?v=OXTn52kL0Yo
Crypto and Securities Laws (Sara Hanks) https://www.youtube.com/watch?v=ocrqgkJn4m0