(j3.2006) Finalization ordering question

[Michael Ingrassia]

Bill Long says
>my original question is unchanged - why do we force 
>the implementation to finalize all the fe's before starting on the ft's?

Right, it may be that this question needs to go back to the original
designers.  Kurt's paper 99-108r1 in meeting 148 has the parent component
being finalized interleaved by element as you suggest, if I'm reading this
right:

>(2)  Each element of the object is processed as follows:
>     
>     (a)  Finalization is performed (recursively) for the explicitly declared
>          component subobjects of the element, in the reverse of the order those
>          components were declared.
>     
>     (b)  If the data type is an extended type, finalization is performed
>          (recursively) for the parent component subobject of the element.

which means we changed it either deliberately or inadvertently later.
I'll see what if I can find anything else in the record.  I wasn't on the
committee at the time, so I have no first-hand memory of this stuff.

	--Michael I.