On the one hand, an embedded size makes sense: I can hop through collections. Except no, it doesn't make sense, because collections are going to be delta-encoded against the first point of the first element, so I have to read every coordinate up to my feature of interest in order to understand the coordinates. Something has to give. Either groups/collections have to commit to including an absolute coordinate at the start of each sub-geometry or size has to go.
On the one hand, an embedded size makes sense: I can hop through collections. Except no, it doesn't make sense, because collections are going to be delta-encoded against the first point of the first element, so I have to read every coordinate up to my feature of interest in order to understand the coordinates. Something has to give. Either groups/collections have to commit to including an absolute coordinate at the start of each sub-geometry or size has to go.