Tracking `subsumes` across major version changes

[tychonievich]

In reviewing how we mark when a v7.1 structure subsumes a v7.0 structure, we noted that some (but not all) v5.5.1 structures are subsumed by v7.0 structures. Adding that information would require two steps:

1. Identifying the subsumes relationships (the automated process used for minor versions will not work for major versions that might include breaking changes).
2. Adding something to the build/extract-yaml.py tool so that the corresponding subsumes entries are added to the v7.0 YAML files.

Doing this will also be something we need to consider when moving from v7.x to v8.0.

[tychonievich]

I propose creating a major-version-subsumes.tsv file we manually edit and entering into it known cases where a subsumes relationship exists between major versions.

Not that we can have "old subsumes new" when something was removed, "new subsumes old" when something was added, and both when no change was made. I didn't find an example of "old subsumes new" in my scanning of v5.5.1 but if v8 keeps the ADDR structure it might be a case of that, removing the deprecated ADR1/ADR2/ADR3 substructures.

I propose a simple two-column spreadsheet, where an unchanged situation would have two rows. But I've struggled with what to name those two columns. From the YAML format spec, A subsumes B if replacing B with A (without changing any payloads, substructures, etc) results in no change in meaning at all. A is thus a more general (or identical) version of B. Based on that, perhaps "general" and "specific" as column labels? Something like

general	specific
https://gedcom.io/terms/v7/NAME	https://gedcom.io/terms/v5.5.1/HEAD-SOUR-NAME
https://gedcom.io/terms/v7/AGNC	https://gedcom.io/terms/v5.5.1/AGNC
https://gedcom.io/terms/v5.5.1/AGNC	https://gedcom.io/terms/v7/AGNC

[tychonievich]

Discussed in steering committee

• The idea of a manually-created TSV file to store this seems fine
• It's not clear if we want "old subsumes new" entries at all. Do v7 URIs subsume v7.1 URIs? Probably not because if they did we'd not have changed the URI. Is there value in helping people go back to older versions of the spec?
• The column labels "general" and "specific" are not great. Maybe
  • "new" subsumes "old"
  • "subsumer" subsumes "subsumee"

We welcome input on these questions. We plan to revisit this in the future and eventually move forward with this.

[tychonievich]

#774 added the tooling necessary to track this; it now remains to populate subsumes.tsv with structures that did not change or were expanded upon in a backwards-compatible way between 5.5.1 and 7.0.