It has been ages since the last time I blogged. Here it is "Those concepts in the geospatial field that have cost us millions of $$$":
https://erouault.blogspot.com/2024/09/those-concepts-in-geospatial-field-that.html
Those concepts in the geospatial field that have cost us millions of $$$

Every domain has its baggage of concepts, which at first sight don't appear to be that terrible, but which are in practice. Let's start with...

@EvenRouault empty geometries are insane 🤬

@nyalld @EvenRouault And *typed* empties are doubly insane.

It's like distinguishing between 32 bit and 64 bit NaN.

@sgillies @nyalld I've always dreamt about being able to use the degrees of freedom left by all those possible bits in a NaN representation to actually store useful information in them. That's a bit like quantum physics with Vacuum energy or this old saying "[Redacted entity] created not-a-number, saw this was (not) good, and from it/them created numbers"
@EvenRouault @sgillies @nyalld R's NA for floating point numbers (distinct from NaN) actually does that; IEEE 754 however doesn't mandate its propagation (but current processors seem to do so); see for a longer discussion the NA section in https://blog.r-project.org/2020/11/02/will-r-work-on-apple-silicon/
Will R Work on Apple Silicon? - The R Blog