GeoArrow is undoubtedly a superior format to GeoJSON, yet GeoJSON remains the dominant standard in web browsers. I'm hoping we'll reach a point where using GeoArrow is just as easy and seamless as using GeoJSON. I hope my talk helps to figure out what's still missing in the ecosystem.

Totally (as long as the ranges are small)! Once the ranges get large Parquet tooling is more likely to use multiple threads to do the fetching (not a limitation of the format, just a limitation of the average tools used to read fgb).

There are a number of plonked files at geoarrow.org/data served via http that all can do this kind of partial read. Any web server that can handle a range request will do (also required for FGB/COG)!

I'll write a post soon specifically on #rstats, but this week's post includes a full example of how SedonaDB can do this just as well in R as it can in Python!