RE: Retrieve Program References

Dennis,

I don't want to be smug here, and with all due respect, but we have been
focussing don this precise issue for some 25 years. I am sure you can
appreciate that in that time we have reworked, discussed, redesigned,
optimised, rewritten this aspect thousands of times collectively.

Based upon this experience my advice would be: don't over engineer how you
extract in the first instance. Build a mapped data set of all references up
front even allowing for redundancy; having too much data is easier to handle
than not enough, the relative disk requirement is irrelevant and its cheap.
Keep it simple but make it comprehensive. This will ALWAYS pay dividends as
your requirement for using these references becomes more complicated and
sophisticated. How else would we programmatically extract business logic,
data models and complete application specs without this fundamental data in
place. In work we did with IBM Rochester over a number of years we looked at
how we might instigate these type of dynamic extractions as needed, they
even built some new API's to facilitate what we do for some of the big
iSV's. The net result was: extract it all up front and then get clever using
the results.

I am intrigued as to what you might be doing with the results of the
extraction. I probably have an opinion on that too if you can bare it.

Stuart