Roadmap: which privacy metrics and input formats should deidproof add next?

[cognis-digital]

Opening a thread to collect metric and format ideas for deidproof. A few on the table (some already in the issue tracker):

1. Entropy l-diversity / t-closeness -- distinct-l alone is vulnerable to skewness (a "2-diverse" class that is 99% one value). A --l-metric {distinct,entropy} flag and/or a t-closeness check comparing each class distribution to the global one.
2. IPv6 detection for S15 -- the value scanner currently only catches IPv4.
3. More input formats -- Parquet / newline-delimited JSON / a --stdin reader, so the same k/l/Safe-Harbor gate works on non-CSV exports.
4. Column-level suppression suggestions -- given a failing dataset, recommend which QI to coarsen next (the report already surfaces the smallest equivalence classes; this would turn that into an actionable hint).
5. Configurable header keyword packs -- let sites add domain-specific identifier column names (e.g. internal accession-number schemes).

Which of these would help your workflow most, and what did I miss? The design constraint is that the core stays standard-library-only and fully offline.