Observations about Serendip's Information Structure
- Information units (web pages, directories) stay in fixed locations.
- Links to and from information units change dynamically, but most often when
the information unit is located "outside", i.e. not under local
- Existing local links among information units tend to stay stable, but additions
to their number grow as an exponential function.
- Information units which aggregate other information units to perform routing
and directory functions are created when some threshold number of information
units is reached (such as n=10). Depth of storage tree equals 2 for nodes
~= 1,000, equals 3 for nodes ~= 10,000, equals (n-1) for 10n nodes.
- Some small number of information units (web pages, directories) stop growing
and/or become orphaned.
- Multiple paths to information units are available; finding the most efficient
path is an instance of the travelling salesman problem. Centralized routing
mechanisms (e.g. information unit ordered by reverse chronology, information
unit ordered by physical topography, search engine with hash table) become
more necessary for finding information with larger scale (depth=3).
- Random discovery mechanisms (e.g. randomizer logo) have a role in interactions
with the system as a whole.
- Choices for locating information storage can be somewhat non-obvious or
arbitrary when the information matches criteria in more than one category.