> The vtable approach used by FlatBuffer seems harmful for
> streaming. I don't think the SBE people would have used
> such a design.
If the vtable and fixed-size fields are placed together with variable-length fields (e.g. byte blobs) at the end, then decoding can make better use of the processor cache than if the fields are all mixed together.
But the idea with the vtable is that many instances of the same "table" (record) would share a vtable. So the vtable would presumably never appear inside the table instance, I think.
