I wrote: > (although, as you can > see, it makes it impossible to make the type non-abstract by actually > providing specific bindings, so you should probably rethink your > approach anyway). I see Malcolm and I are both typing in Las Vegas simultaneously. Anyway, this is what he said as well: In practice your goal will not work regardless of whether NAG issues an error message or warning or nothing and just lets you endlessly extend abstract upon abstract type. Best, Aleks