> Doesn't size also need the header word to determine which fields are > pointers for the traversal? Yes. > (But, I guess that information must be around > somewhere. Right. The mark bit, OTOH, must be per object.