I'm trying to optimise calls to a lot of vdb points methods where I know the attributes are expanded and uncompressed. In almost all cases the best solution would be to retrieve the attribute buffer directly. One example is performing in-place updates of attributes values or being able to use methods which take a pointer to the start of a contiguous array and their sizes. This is similar to how LeafBuffer's .data() can be leveraged. danrbailey thoughts?