The idea of clickstreams are not new. The following describes a set of features that could make up a browser implementation of clickstreams or an extension. The coupling of clickstreams with ant theory for mapping routes could create a new way to browse through a user's history, automatically populating forward and back buttons based on probability that a user at a particular place on the internet, will navigate to a linked page.
Store users' paths through the internet
A digraph of nodes representing how a user got from point A to point B by way of all nodes in between.
A user's link paths through the internet can be stored locally.
Either as a function of synchronization or as an opt-in sharing mechanism, clickstreams can be uploaded to a central application for group sharing. Clickstreams would have to be anonymised and inscrutable if they're to be stored on a server.
Update nodes in a path with a "heat" value
Nodes on the internet graph can be marked with a value based on visitor frequency. A floating point value from 0 to 1 represents the relative heat of any particular node. Values are incremented logarithmically with "1" being an asymptote. As the clicks that took a user to a node on the path age, the heat is decreased. Paths can change over time and will reflect the "current" trends in internet navigation or be able to map popular routes at a given point in time.
Next / Forward
From any node on the internet, a collection of "next" nodes can be inferred based on the heat values of the adjacent links.
Local or Global
Local view or a global view (or some mixture of the two) depending upon preference or whether or not the user has opted to participate in the shared application. Users could choose to mix local or global paths using a weighting factor to increase or decrease the weight of local paths.
Paths vs. History
One augments the other. Cache settings and implementation need not change to make use of this feature. Paths are stored separately, but can take advantage of cached history for faster load times. As paths age and expire, corresponding history data has already been cleared from the cache. Only current "hot" paths are retained and cached locally.
Clickstreams become a basis for prefetching and caching next nodes. The fetch can occur quietly in the background, based on the likelihood of a user to travel to the next node in the clickstream. Heat values correspond to probabilities.
Peer to peer sharing of clickstreams (opt-in, invite only, no central server communication) could allow friends and networks of friends to share clickstream data. An interface could be superimposed to show which of your friends or colleagues visited a site and integrate with Places or Bookmarks to show how these users tagged the pages.