Ontology-Driven Observability: Building an End-to-End Knowledge Graph at Netflix

Commenters debated Netflix’s push to use an ontology-driven knowledge graph for observability, with some endorsing the idea as a way to formalize meaning and improve system-level communication and others warning of practical pitfalls. threecheese argued machines will “dictate” ontologies and normalize away nuance in favor of token efficiency, while delichon pointed to the “bitter lesson” as a rationale for ontology-based development. Several participants, including smartmic and 1970-01-01, expressed enterprise-grade skepticism: smartmic described long, costly experiments with little widespread uptake and 1970-01-01 demanded concrete evidence that this reduces latency or delivers measurable engineering value. dominotw dismissed the effort as overengineering.

Other commenters focused on definitions, tooling and learning resources. antonvs insisted “ontology” is broader than RDF/OWL triples, and rapnie highlighted developer resistance and the complexity of the linked-data stack (citing ActivityPub’s partial JSON use as evidence). shervinafshar, bendriv and threecheese suggested practical readings and tools (e.g., the SEP entry, McGuinness papers, LinkML), and CharlieDigital asked for clearer articulation of business outcomes. Views divide between those who see ontologies as necessary for shared meaning and systematizing observability and those who see them as complex, under-supported, and unproven in delivering operational value.