Blog: announce 2.2.0 and 2.3.0

[lambda2]

Covers both releases. 2.2.0 never got a post, and a gap in the series would read worse than one post spanning two versions.

The through-line

2.1.0 introduced provenance — recording which source each value came from. These two make that recording mean something: deciding which source wins when they disagree, and letting readers watch the dataset improve day by day.

What it leads on

Taxonomy, because it is the change most likely to affect what people actually read. Whether a name is accepted or doubtful, who authored it, when it was published — all still on last-write-wins, so a status from Kew could be replaced by one from an aggregator whose taxonomy is assembled automatically, with nothing recorded. Explained without naming and shaming any particular source.

Also covered

• Arbitration rules in plain terms: ties never overwrite, older values hold a rank of their own, implausible values are refused and kept on record, closed vocabularies
• source_record_id on facts, with a concrete JSON example
• Month and part fields now readable ("mar|apr|may" rather than "28")
• The daily evolution view, including why we measure filled cells rather than an average completeness score
• The three smaller fixes users would notice: boolean filters that no longer guess, the unknown image key, cultivation fields expected only of cultivated species

Register

Same as the previous posts — says the traits are largely empty rather than dressing it up, and closes by asking which fields people actually need, since lookup frequency is a poor proxy for that.

Build green, page generated at blog/2-3-0-released.