GEOGRAPHIC BOUNDARY DATA: PROVENANCE AND TERMS

This notice accompanies geographic_boundaries.db and its JSON manifest.
Bokeh's software license is in the distribution's LICENSE.txt. The following
source-specific information applies to the boundary data.

US STATES AND COUNTIES

These are re-encodings of existing Bokeh sample data, not newly acquired
external datasets. Inputs: US_Regions_State_Boundaries.csv.gz and US_Counties.csv
from bokeh/bokeh_sampledata revision:
4195d5e5c3094ef578be168778c7e9501138f921
https://github.com/bokeh/bokeh_sampledata/tree/4195d5e5c3094ef578be168778c7e9501138f921/src/bokeh_sampledata/_data

Geometry source identified by comparison: U.S. Department of Commerce,
U.S. Census Bureau, Geography Division, TIGER/Line Shapefiles. Verified against
the official 2009 current-state/current-county national releases:
https://www2.census.gov/geo/tiger/TIGER2009/tl_2009_us_state.zip
https://www2.census.gov/geo/tiger/TIGER2009/tl_2009_us_county.zip

Source credit: U.S. Census Bureau, Geography Division.

The XML metadata inside both archives states that the TIGER/Line shapefiles
are not copyrighted and are free to use in products or publications, with
acknowledgement to the Census Bureau as source. TIGER/Line and Census TIGER
are Census Bureau registered trademarks. These terms permit redistribution
alongside BSD-3-Clause software. Retain the Census source acknowledgement.
The boundaries are for statistical purposes, not legal land descriptions or
determinations of jurisdiction, ownership or entitlement. Source coordinates
are NAD83 longitude/latitude. This conversion does not reproject them.

Comparison evidence: all 51 state identifiers and all 3,233 county identifiers
occur in the 2009 reference. Every one of 14,134 distinct state vertices and
506,908 of 506,926 distinct county vertices matches a vertex of the respective
Census feature within 0.0000051 degrees on each axis. Counts deduplicate points
within each feature and use the final source row for duplicate county FIPS.
The 18 unmatched county vertices occur in 11 features. Eight belong to the
coarse Caldwell County, Missouri (29025) polygon. This is strong evidence of
simplified, rounded Census geometry, not a byte-identical original export.

The state CSV is byte-identical to the file committed to Bokeh on October 10,
2013 (commit 993f6709b50e4d8ec2e332aae5d21c6246e836b2). The loaders were added
on October 7, 2013 (commit 5dd06b424072da46d910aacf282e59bef314d7fc), when
counties were downloaded from Bokeh's own S3 bucket. Google Fusion Tables is
a plausible intermediary, but an exact export/table attribution was not
recovered. The exact original release year, simplification procedure and
origin of the additional region/color metadata remain unrecorded. 2009 is the
verified matching reference, not an asserted exact acquisition vintage.

The Bokeh repository's BSD-3-Clause license covers its distributed software
and contributions. It is not by itself evidence of the original data rights.
The Census source metadata supplies the independently verified data terms.

Conversion preserves polygon interiors and repairs invalid polygon geometry,
retaining polygonal components and discarding collapsed line/point remnants.
Duplicate county FIPS identifiers retain the final source row. The JSON
manifest records original input hashes and every repaired identifier.

REPRODUCIBILITY

See tools/README.md in the source repository for checksum-verified input
retrieval and deterministic regeneration using the committed Pixi lockfile.
The manifest records counts, input/output checksums and generation versions.
