I didn't mention this in my original post, but I'm putting the subs in seperate modules. Which is part of the reason for the original questions - the other one being the recent node on limits of nested subroutine calls.

I'll be processing fairly large ( for me ) data sets, from 85MB up to 200MB or more, in one pass. Nesting seems to be the best path I can think of to break the processing up into smaller chunks of related data.