RE: Object ownership?

>>The spec could be read to imply that Node insertion passes ownership to
>>the containing node and removal/replacement returns ownership to the
>>caller. Is this a reasonable interpretation?
>
>That might be a reasonable approach, but I know that *I* never wrote
>anything into the spec with this thought in mind.  Can you point to a
>particular bit of the spec that needs clarifying?

I think it is implicit that node insertion passes ownership to the
tree/parent-node regardless of the language or memory management being used.
It is unreasonable for a DOM user to maintain references to all nodes
inserted into a given tree such that the user is required to explicitly
delete/release all the inserted nodes when the tree/parent-node is
deleted/released.  When the tree/parent-node is deleted/released, assuming
there are no remaining references to any children, all those children will
be inaccessible (because the only way to get to them was just
deleted/released) and the user will have no way to otherwise delete/release
them.  The tree/parent-node must, then, delete/release them to not have
working set bloat or a memory leak. 

Actually, as I write this, I'm wondering if it is possible to implement the
DOM without using a reference counting model.  Above I said that the parent
node must release/delete any children.  If a non-reference counting model is
used, how is the parent to know to delete a child or not?

Perhaps I have spent too much time writing apps which use reference
counting!  Can anyone address a memory management scheme (other than garbage
collection) for the DOM where reference counting is NOT used which addresses
the example above?

- Eric

Received on Wednesday, 27 May 1998 02:11:31 UTC