Work in the Archimedes Project involves two sorts of data: (1) data that are automatically generated and may be regenerated with ease, and (2) data that cannot be generated without significant human effort. We call the first category of objects transient or volatile, and the second permanent or non-volatile. For example, automatically generated morphological data are considered transient, whereas the files corresponding to primary source texts are considered permanent.
The Archimedes Project ontology comprises all permanent objects in the project, and only those objects. The ontology is not flat, but rather is structured in terms of hierarchical relations between objects. Primary objects are independent of any other object in the ontology. For example, the text of a specific edition or manuscript is a primary object. First-order derivative objects depend directly on at least one primary object in the ontology. We distinguish between first-order derivative objects that depend on exactly one primary object (e.g, photographic images of the pages of a text edition) and those that depend on multiple primary objects (e.g., text matching files). Second-order derivative objects depend directly on derivative objects, but they depend only indirectly on primary objects. For example, figures depend directly on the image of the page from which they are taken, which depends in turn on the source text. Also, any file that relates first-order derivative objects (such as one that associates instances of technical terms in primary texts with more general concepts) is a second-order derivative object.
Every object in the Archimedes ontology is assigned a unique locator. This locator provides a stable reference to the object, and should never be changed or reassigned. We describe the conventions for such locators in section 3.