Technically they exist in HTML+HTTP via the (mispelled) `Referer` header. It sucks that cross-site Referer sharing is seemingly being phased out by browsers in the name of user-privacy (as referer URIs do often contain PII), but there are more cynical explanations afoot which I won’t go into.
Another approach is having a central link-repository that works as an overlay on the fully distributed web - this could be provided by search-engines.
Another approach is a distributed link database. I’m sure some clever DHT system could be used to allow practically anyone to participate without needing nodes to hold a copy of the full database (a-la Bitcoin), or even one that allows directly-relevant nodes (e.g. a website’s own node) to stores only it’s own bi-di links and not waste resources on unaffiliated content.
Of course, all of this assumes that spam, trolling, griefing, etc don’t exist :)
For all the talk WWW gets about being a distributed system... it isn’t: it’s more of a federation of systems united only by their common use of HTML+HTTP and (hopefully) RESTful resource architecture. A true distributed WWW would be some horrible mix of IPFS, BitTorrent and Ethereum. I think that’s how Roko's Basilisk will come into existence...