Different row heights and user adjustable row heights will be similar to the existing code for columns. The main difference being that large tables normally incorporate more rows than columns so speed and memory consumption really become a part of the implementation.
In keeping with tradition, this new resource will be allocated on an as need basis so existing applications will still maintain their performance.
Different row heights will be difficult to specify via a resource and I envisage that the widget will handle most of the computation, whether it's a pixmap that is being displayed or a different font or more than one line of text. I feel that these parameters cannot be programmatically assigned and will remain internal to the widget.