>Fabio,
>
>My bad Fabio, I missed out on the word "table", which is as you have said, does not get erased when private DS is destroyed.
>
>This method, together with using XML, are the two methods I will be using.

And note that the speed of XML drops squarely with its size. It being a text file, it just has to be parsed; the speed of parsing is beyond our control. So for larger datasets you may actually be better off with a temp table, or a collection of scatter name objects (and the dangling DS should be possible to solve by tablerevert()ing every cursor it has).