An early experience in connecting manuals to software failed because subjects weren’t used and described consistently. Two outcomes of this failure were DocBook and Topic Maps, initiallly implemented in HyTime, but quickly afterwards in XML, producing XTM (see topicmaps.org). XTM was adopted into ISO 13250. The revisions to the current version should be mostly completed this year or next.
A diversity of identifications is a given: there won’t be a single identifier that everyone will adopt. This is an old problem in computer science that goes by lots of different names, like record linkage, entity resolution, etc. The “my identifier” method assigns a unique identifier to each thing: but you have to trust their judgment about what’s split and joined, and the original distinctions have gone away.
In Topic Maps, a subject is represented by a “topic”. A separate mechanism deals with relationships (“associations”) and “occurrences”, an instance of a subject. Subjects have identifiers and locators (like a URL). The Topic Maps Reference Model is different: an abstract model where subjects are represented by proxies. Identification matters because it defines what we can talk about. Topic maps give us a way to integration information across separate databases. Subject-centric computing is another old concept. We need some basis for disclosing our rules for merging: that way it can proceed bottom-up.