What Is An Information Resource?

We can understand a resource as a relation between one or more identifiers and one or more concrete representations, themselves related to one another by content-preserving transformations.

A Candidate Typology

To classify the properties of information resources, let us consider two predicates as perpendicular dimensions:

Examples of resources at the different intersections of the two predicates.
Sampled Computed
Opaque A document that you wrote, a picture that you took—or anybody else, really A zip file, a document converted from one format into another
Transparent Data collected from a form, or retrieved from an API call Aggregates or other functions over sampled—or even other computed—data

Provenance Dimension: Sampled vs. Computed

The first dimension considers how the resource was obtained.

An information resource is sampled when it comes from outside of the information system under consideration, and therefore the system cannot have an account of how the resource—or rather its representation—was produced. Sampled resources must therefore be preserved in case the source becomes unavailable and they can't be obtained a second time.
An information resource is computed when a known process is applied to zero or more resources already under management within the system, generating, manipulating, combining, or otherwise transforming them. Computed resources can, by definition, always be recomputed, thus making them in principle more disposable than sampled resources, although in practice the computation always incurs a cost.

Interpretation Dimension: Opaque vs. Transparent

The second dimension considers the interpretability of the resource by the information system that houses it, including the addressability of the resource, and its propensity for isomorphic representations.

An information resource is opaque when there can be said to be a single, unitary, concrete, literal string of bits which defines the resource's canonical, authoritative representation. All available transformations will lose information which will be evident when their inverses are applied. Furthermore, since every opaque resource originates as a literal string of bits, it can be said to have multiple readings, alongside the intended semantics of the data format.

An information resource is transparent when it has no canonical representation whatsoever: when any one representation is perfectly as good as any other in terms of its informational content. We call it transparent because the information system can see straight into it, and therefore store, retrieve, and manipulate the content to arbitrarily fine grain.

Note that a resource initially considered to be opaque may be discovered to be transparent through a round-trip parsing and serialization operation. Examples of this phenomenon could be a CSV file ingested into a database and spat back out again, producing, bit-for-bit, exactly the same file, or the same thing with a Markdown file which can be turned into HTML and then stripped back down into an identical original.

Why Care About This Stuff?

I consider these to be fundamental structural properties of information resources, which themselves constitute a suitable basic unit of an information system. These structural properties are particularly instructive in helping design such systems.