According to some research, webpages are only available on average for 100 days. This can make linking to information for a long term unreliable. In some cases, P2P can be designed so that pages don't disappear if they're made no longer available from a single source, because the files are mirrored across the network.
Wikipedia entry regarding content-addressable storage. This is an alternative to domain names, where files are linked over a distributed network via a unique hash. The hash is used to discover seeders of the file.
A technique used in discovery of peers in a network. However, I wonder if a similar approach to information redundancy would be to provide multiple links to locations of similar information.
This lib is a set of tools that can be used to build P2P systems.
A GNU project that provides a set of tools for building a new Internet stack. Defines a naming system as an alternative to DNS.
Further background to the motivations behind GNUNet
An active P2P hypertext implementation. Can also extend into local and offline networks. Uses a concept called Activelinks that utilise images to store the link information, but text links can also be used.
A networking tools stack built by Freenet. Seems similar to GNUNet.
A P2P distributed file system implementation.
Distributed data store. Similar conceptually to IPFS but doesn't appear to have as much adoption. Based on hypercore protocol.
A stack of tools to build P2P applications.
Project is no longer active. A browser that is used with Dat.
File exchange protocol, distributed and decentralized. Magnet links could be used to link to files. Haven't seen this used specifically for hypertext, but possibly could be.
Wikipedia entry for magnet links.