Thu Apr 8 12:58:58 EDT 2010

Representing links

Is there anything useful to know about graphs in general from my
perspective?  I.e. I don't really care about structure that much, more
about representation and direct connectivity. i.e. neighbours of one
page, not neighbours of neighbours.

It seems that the basic idea is closure.  If a document contains a
link, this can either be pointing into a closable neighborhood, or to
an outside resource.

We'd like to keep track of both: 

  - represent neighborhood references such that wrong links can be
    detected statically.

  - gather open links to enable link checking and queries over links

Let's have a look at this[1] zipper-based rep.  It mentions that in a
functional representation, for every cycle there has to be a point of
decoupling which can be solved either by mutable cells or a finite map
combined with unique identifiers.  I suppose this is true if the graph
can't be constructed in a single recursive `let'.

[1] http://www.cs.tufts.edu/~nr/pubs/zipcfg-abstract.html