You're right, we should delve into a comparison more with respect to prolly trees. We actually have a lot of experience with prolly trees, and have found, in practice, that you need to do a lot of the things that folks like dolt have had to do to make them work nicely. Whereas with G-trees, the basic implementation turns out to be quite nice (and extremely easy to reason about).
One of the biggest benefits of G-trees in my mind, is their ease of implementation. Additionally, we did a lot of work to explore their statistical properties, which doesn't exist for prolly trees (though in hindsight, we have done this, so should probably write it up formally).
One of the biggest benefits of G-trees in my mind, is their ease of implementation. Additionally, we did a lot of work to explore their statistical properties, which doesn't exist for prolly trees (though in hindsight, we have done this, so should probably write it up formally).