Content-addressing is unusable with an index or database that describes each piece of content. Since IPFS is fully content-addressable, nothing can be done with it unless you have a non-IPFS index or database, or an internal protocol for dynamic and updateable links.
The IPFS conceit made then go with the with the second option, which proved to be a failure. They even incentivized the creation of a database powered by IPFS, which couldn't be more misguided.