The idea of primitives or atomic elements is always popping up from time to time. From a practical standpoint, this concept is very useful when ontologies are used to describe implementations of systems (such as databases, services, real systems, etc.). The primitives should be one-to-one mappings to the primitives used in these systems. The reflect scope, resolution, and structure and capture the reigning constraints as well. Such ontological representations are very valuable to discuss and reason about composability or interoperability of solutions. These primitives are primitives of the conceptualization that underlies the implementation.
I also agree that it is necessary to agree on ontological primitives, as they allow a standardized way of representation. Ontological primitives can unambiguously describe primitives of conceptualization.
However, these primitives of conceptualization are modeler or system designer decisions and to not reflect physical or logical necessities. Ontological primitives are the syntax useful to describe them.
If there are primitives of meaning in the real thing is a philosophical question, that is of secondary interest to the engineer, as he is always using a conceptualization of the real world referent. Following Odgen's idea of the semiotic triangle, we can even make the case that we actually all use conceptualizations to refer to the real world referent. Each conceptualization is a facet to the real thing, and uses its own artificially introduced primitives. Each new facet broadens our understanding of the real thing ... and this may result in digging deeper and deeper and adding more accuracy, higher resolution, and other insights.
In my world view, primitives are real for each conceptualization, but they may not be real for the real thing, which we never can grab without conceptualization anyhow.
All the best
Andreas Tolk, Ph.D.
Engineering Management & Systems Engineering
Old Dominion University