Hi Frank,
I hope we can have a more fruitful exchange on this topic, but here are some preliminary comments on requirements in EA.
For most projects, I create a "Project definition and requirements" package as the first package in my EA projects. I'm currently experimenting on having this requirements package as part of a more general "Subject matter area analysis" package, together with the area's org chart package, the processes package and the domain structural model; but this not essential, and is only a matter of how you organize your project tree.
In this project definition and requirements package I create a diagram with the relationships between requirements and goals. I got the idea from Eriksson and Perker's "Business Modeling with UML" (see "Goals", under chapter 3: "Modeling the Business Architecture"), though I add requiriments in addition to these author's quantitative and qualitative goal diagrams.
Why is this useful? 1. Because I get a visual diagram with a model of project goals and requierements; 2. EA can generate a pretty nice RTF from this package. The RTF contains a description of each goal and each general requirement, the org-chart, the process diagram and the domain structural model (formerly called "Logical business model", and that contains the "real world" objects you first discover in the subject area; sort of a pre-structural-class model). You can add, also, a preliminary technical architecture package, which will also be part the RTF (or, if you preffer, of the HTML) doc.
In the diagram, all major requirements and goals are associated with the project's central goal; for instance: "Design, develop, test and put into production a system for capturing, storing, concentrating and analyzing individual and community data, in order to support decision making by municipal and state authorities". The general requirement to put this systems into production "depends on" adequate testing, which "depends on" building the system, which "depends on" design... and so forth. It is simpler to display these relationships visually that to explain them in the text.
The RTF I generate with these packages (including diagrams) is the heart of the project's scope document, which is the most important specification you need the customer to agree on --in order to mitigate scope creep-- before you embark on further analysis (for instance, before you embark on the use case model) or design. It is also an extremely useful document to reference in a formal contract.
If I model and base-line goals and requirements this way in EA, then I do not need to write and maintain a separate scope document; instead, I have a single place where I introduce new general requirements and goals, objects in the org chart, etc. In other words, I can control project scope (largest single source of project headaches) in a the same project file where analysis and design are contained.
I guess we can add this scope document deliverable to the PUP (Phil's Unified Process, which I propose to be built by EA forum members as our alternative to the proprietary Rational Unified Process; and, believe me, people here in Mexico will get a kick out our process being called PUP).
Jaime Gonzalez