Thoughts on Intercommunication: Links

A seemingly-fundamental type of data which should be representable in any ideal network is the link.
It's common for programming systems to have implicit links with many different uses, and to restrict
them to what are commonly called ``references'' that have fewer fundamental operations than the more
primitive ``addresses''; links aren't a real type, however, revealed after trying to represent them.

Summaries are most commonly used to represent links in many systems, but they use trapdoor functions
to approximate unforgeability, and trapdoor functions belong not in an ideal network where they will
be broken after a time, in all likelihood.  Nothing that could possibly be replaced or scorned ought
to be in the unmovable foundation of an ideal network.  A network ``address'' is a different kind of
address and isn't generally unforgeable, since that rather flies in the face of delivering messages.

The Domain Name System features linked data used for simple compression, represented as indices into
the payload, but they're clearly nothing more than simple integers meant to index the message array.
At the level of the Border Gateway Protocol, and the Autonomous System organization of the Internet,
one realizes even addresses to be illusory things.  Generally, getting links is logarithmic at best.
Regardless, the concept of association is one more useful than the relatively vague notion of links.
Since anything can be an association to anything else, trying to bake in one form is worthless work.

When one realizes the inherent nature of the ``address'' to be illusory, merely one association into
arrays made implicit and merely one type out of many possibilities, one realizes associative data to
be much richer than many believe.  At the level of representation, an address is but an integer, and
must have a context in which to be useful.  Similarly, any other association in the data is at least
as useful, if not moreso.  Most often, links can be avoided entirely: A linked list, a doubly-linked
list, and an array have identical representations when used to hold an ordered collection; that type
to choose and to use when decoding a message is best left to a higher level.  Any associations after
that most common case probably ought to be made explicit, and task-dependent, such as representating
data associated with a symbol using whichever types work best rather than which are more ``innate''.

Computing is nothing but artifice, and confusion over which artifice be more ``inherent'' is deadly.
.