{"updated":"2026-11-04T19:51:00Z","content_md":"I have ranted against the brittleness and unusability of formal ontologies [Politics and Pragmatism in Scientific Ontology Construction](https:\/\/ammdi.hyperphor.com\/Politics-and-Pragmatism-in-Scientific-Ontology-Construction) but for my sins, a friend is trying to get me to work on an OWL ontology for semiconductor manufacturing processes (admittedly, a domain that they might actually be suited for). Ontologies are standards and sort of protocolish, although they seem more static than a protocol.\n\n![[4y2w44y42hpnwnchhqktee8c3x]]","changelog":[{"version":1,"at":"2026-10-27T00:52:07Z","note":null},{"version":2,"at":"2026-11-04T19:51:00Z","note":null}],"transclusions":[{"id":"4y2w44y42hpnwnchhqktee8c3x","version":2}],"created":"2026-10-27T00:52:07Z","page":"f\/47j2g7v177fzvzb0y0h8zjrfs4\/","content_hash":"sha256:d75c345d6f58e70a113a065e39e65cf58d052157bd128de73f10261350ab7681","id":"47j2g7v177fzvzb0y0h8zjrfs4","kind":"thread","origin":"https:\/\/ammdi.hyperphor.com\/blyg\/","content_html":"
I have ranted against the brittleness and unusability of formal ontologies Politics and Pragmatism in Scientific Ontology Construction<\/a> but for my sins, a friend is trying to get me to work on an OWL ontology for semiconductor manufacturing processes (admittedly, a domain that they might actually be suited for). Ontologies are standards and sort of protocolish, although they seem more static than a protocol. <\/span><\/div>
Ask Claude<\/span>
    \n
  • This is a rich set of tensions worth unpacking. Let me engage with both the "is this a good domain fit" question and the "ontologies vs. protocols" framing.<\/li><\/ul>
      \n
    • <\/li><\/ul>
        \n

      • Semiconductor Manufacturing as a Plausibly Good Domain for OWL Ontologies<\/h2><\/li><\/ul>
          \n
        • Your instinct that this domain might<\/i> actually be suited to formal ontologies is worth taking seriously. The core reasons:<\/span><\/li><\/ul>
            \n
          • <\/li><\/ul>
              \n
            • Semiconductor manufacturing involves numerous complex steps and processes requiring extensive knowledge for effective execution, where in each step having an expert specialized in that particular field to guide workers is crucial.<\/li><\/ul>
                \n
              • That's precisely the kind of structured, hierarchical, relatively stable<\/i> domain knowledge where formal ontologies earn their keep \u2014 unlike, say, social science concepts that are inherently contested.<\/span><\/li><\/ul>
                  \n
                • <\/li><\/ul>
                    \n
                  • To address knowledge structuring challenges, some groups have collaborated with semiconductor experts to develop ontologies that systematically structure the entire semiconductor manufacturing process, constructed using a top-down approach, dividing the field from general to detailed levels, sub-levels, and specific processes.<\/li><\/ul>
                      \n
                    • The interoperability use case is concrete:<\/li><\/ul>
                        \n
                      • at the level of manufacturing operations management and control, ontology-based models following standards like IEC 62264 have been proposed to address interoperability issues in production-control systems, and Process Specification Language (PSL) ontologies with formal representation of manufacturing processes can help exchange process information automatically among a wide variety of manufacturing applications.<\/li><\/ul>
                          \n
                        • So there are real prior efforts and a genuine interoperability problem to solve \u2014 not just ontology-for-its-own-sake.<\/li><\/ul>
                            \n
                          • <\/li><\/ul>
                              \n

                            • Ontologies as Standards: More Static Than Protocols?<\/h2><\/li><\/ul>
                                \n
                              • Your intuition that ontologies are "more static" than protocols is mostly right, but the distinction is worth sharpening:<\/li><\/ul>
                                  \n
                                • Protocols<\/b> govern interaction<\/i> \u2014 they define message sequences, handshakes, state machines. They have to be dynamic almost by definition because they mediate live exchanges between systems.<\/span><\/li><\/ul>
                                    \n
                                  • Ontologies<\/b> govern interpretation<\/i> \u2014 they define what terms mean and how concepts relate.<\/span><\/li><\/ul>
                                      \n
                                    • Ontology-driven interoperability produces semantic descriptions of protocols and the data units they exchange; an ontology describes concepts and relationships between concepts in a specific domain.<\/li><\/ul>
                                        \n
                                      • In this framing, an ontology is actually below<\/i> the protocol layer \u2014 it's the shared semantic contract that makes protocol messages interpretable.<\/span><\/li><\/ul>
                                          \n
                                        • <\/li><\/ul>
                                            \n
                                          • Gruber's widely-used definition of an ontology is a "specification of a conceptualization." Ontologies are, at a minimum, sets of terms used in a specific domain, definitions for those terms, and defined relationships between the terms; they can range from simple controlled vocabularies to structurally complex representations employing description logics.<\/li><\/ul>
                                              \n
                                            • The "static" problem you're pointing at is real and known:<\/li><\/ul>
                                                \n
                                              • IoT\/WoT ontologies notably suffer from lack of a grounding, making them require adaptations to interoperate or have proper reuse, with no assurance of semantic grounding.<\/li><\/ul>
                                                  \n
                                                • In other words, even when ontologies are standards, they often lack the mechanical enforcement mechanisms that make protocol standards actually binding.<\/li><\/ul>
                                                    \n
                                                  • <\/li><\/ul>
                                                      \n

                                                    • Why Semiconductor Fab Might Be Different From the Usual Ontology Failure Modes<\/h2><\/li><\/ul>
                                                        \n
                                                      • The brittleness you've argued against in scientific ontology construction typically comes from:<\/li><\/ul>
                                                          \n
                                                        • Contested domain concepts (what counts as a "species"?)<\/li><\/ul>
                                                            \n
                                                          • Stakeholder politics baking in parochial classifications<\/li><\/ul>
                                                              \n
                                                            • The world changing faster than the ontology can be updated<\/li><\/ul>
                                                                \n
                                                              • Semiconductor process steps \u2014 lithography, etch, deposition, CMP, implant, anneal \u2014 are comparatively uncontested<\/i> as categories, physically grounded, and change slowly enough that an ontology version can remain useful for years. The domain also has a clear interoperability payoff: fabs, equipment suppliers, and EDA tools all need to exchange process descriptions, and<\/span><\/li><\/ul>
                                                                  \n
                                                                • automation of planning processes is desirable to cope with frequent changes in semiconductor supply chains, requiring rich communication between different decision-making units, whether human planners or software, and hardware components.<\/li><\/ul>
                                                                    \n
                                                                  • <\/li><\/ul>
                                                                      \n

                                                                    • The Practical Upshot<\/h2><\/li><\/ul>
                                                                        \n
                                                                      • The ontology-as-standard framing is probably the right one for your purposes: treat it less like a knowledge representation exercise and more like a data schema with formal semantics bolted on<\/b>.<\/span><\/li><\/ul>
                                                                          \n
                                                                        • OWL extends RDF with formal logic, enabling machines to infer relationships that were not explicitly stated<\/li><\/ul>
                                                                            \n
                                                                          • \u2014 that inferencing capability is probably the one thing worth preserving over a plain schema, since you can do things like automatically classify a process step as requiring cleanroom class X given its contamination sensitivity properties. The brittleness risk is manageable if you keep the ontology's scope narrow (process steps, parameters, equipment capabilities) and resist the temptation to model everything.<\/li><\/ul><\/div><\/span><\/div><\/blockquote><\/div>","version":2,"media":[],"blyg":"0.3"}