Philippe Rivière

It's not something that we're doing for parameterized pages and resources, _i.e._ you can call `page-foo` if there is a `page-[var].md`, even if that parameter is not listed in dynamicPaths....

We should probably call it labelCentroid, with an ellipseRatio option that defaults to 2 (?). And maybe add an _anchor_ option so that we can align text on the left...

See https://github.com/observablehq/plot/issues/2087#issuecomment-2169156629 for using this as part of a "geo" pointer option.

Related: * [ST_MaximumInscribedCircle](https://postgis.net/docs/ST_MaximumInscribedCircle.html): equivalent of the pole of inaccessibility. * [ST_PointOnSurface](https://postgis.net/docs/ST_PointOnSurface.html): no particular specification re: the quality of the point. It just is on the surface.

A case found in the wild: the binning of integers between 0 and 7, that results (by default) in small bins such as [4 — 4.5]. If the format was...

Having red links (for preview at least) and optionally breaking the build would give us the best of both worlds.

As an additional option, remove the href attribute from red links (https://github.com/observablehq/framework/discussions/1858#discussioncomment-11437041).

the voronoi cell now depends on i and fi (with a test showing facet: exclude)

~~It works!~~ ~~Applying the exclusiveFacets to the voronoi mark only for now (I saw that it didn't work with the voronoiMesh).~~

~~In the documentation there is a voronoi mark with the centroid initializer, and the reindexation logic fails in that case.~~