my point is merely that as written, the article is confusing.
to paraphrase: "hashtables are faster. here's some research showing b-trees are slower than hashtables. why would you ever use a tree? prefix matching, that's why."
which misses the point that the primary reason for using a B-tree as a general purpose database index is not about prefixes, it's about block devices (prefixes is an important reason, but it's not the main one)
if the purpose of the article was to ignore implementation and spinning disks and just look at in-memory performance, then don't mention B-trees. they're an implementation-specific optimisation, and it just confuses.
