inst/prov/fb.prov and inst/prov/slb.prov ship ~1 MB of JSON-LD that is now stale: the records describe releases up to v23.05 hosted on the rfishbase_board GitHub repo, name import_db.R as the generating script (removed in #324), and were produced by the prov / contentid workflow that has not run since hosting moved to HuggingFace and then to Source Cooperative. Nothing in the package reads these files.
The data itself is now published under cboettig/fishbase/{fb,slb}/v<YY.MM>/parquet/ on Source Cooperative, with data-raw/ holding the scripts that generate it (see data-raw/README.md). So the pieces for real provenance exist, they're just not wired to anything.
Worth deciding:
- Regenerate or drop. Either emit fresh schema.org
Dataset records covering the current releases (through fb v26.07 / slb v26.04) or delete inst/prov/ outright. Shipping records that point at dead URLs and a script that no longer exists is worse than shipping none.
- Where it lives. A per-release JSON-LD sidecar next to the parquet on Source Cooperative is probably more useful than a file inside the R package — it describes the data, not the package, and it would not bloat the installed package or need a CRAN release to update.
- How it gets made. If we want this maintained rather than bit-rotting again, it should be a step in
data-raw/import_dumps.sh that runs on every import, not a script invoked by hand.
- Content hashing. The old workflow registered content identifiers with
contentid against hash-archive. Decide whether that is still wanted; if so it needs a home that is expected to outlive the current one.
Not urgent — the data and the package are correct without it. Filing so the decision is recorded rather than rediscovered at the next import.
inst/prov/fb.provandinst/prov/slb.provship ~1 MB of JSON-LD that is now stale: the records describe releases up to v23.05 hosted on therfishbase_boardGitHub repo, nameimport_db.Ras the generating script (removed in #324), and were produced by theprov/contentidworkflow that has not run since hosting moved to HuggingFace and then to Source Cooperative. Nothing in the package reads these files.The data itself is now published under
cboettig/fishbase/{fb,slb}/v<YY.MM>/parquet/on Source Cooperative, withdata-raw/holding the scripts that generate it (seedata-raw/README.md). So the pieces for real provenance exist, they're just not wired to anything.Worth deciding:
Datasetrecords covering the current releases (through fb v26.07 / slb v26.04) or deleteinst/prov/outright. Shipping records that point at dead URLs and a script that no longer exists is worse than shipping none.data-raw/import_dumps.shthat runs on every import, not a script invoked by hand.contentidagainst hash-archive. Decide whether that is still wanted; if so it needs a home that is expected to outlive the current one.Not urgent — the data and the package are correct without it. Filing so the decision is recorded rather than rediscovered at the next import.