Hi, Andy.
When going from analysis to design, it is inevitable that some of your analysis classes will evolve or collapse into new implementation classes or tables. The way this is generally handled is to create separate class models for analysis and design.
During analysis, you create an analysis model with the real-world objects you discover: this is also known as the domain model, but you can call it logical model and use the default packages provided in EA. It is very convenient to keep this model in a diagram of its own.
When going into design, the actual data and user interface models should begin to emerge into their own diagrams. If your are using an RDBMS, it is quite convenient to have all your tables in a data model package (you can copy your analysis classes into the data model with "copy as new"), and name them t_user, t_customer_heading, t_product_detail, etc (so as not to mix their names with analysis objects). Same for w_login, w_main_menu, w_cutomer_header, w_customer_detail....
To give you an example: I have a domain model with country, state, municipality, community, family, person classes, discovered during analysis. In the data model all of these collapsed into t_entity.
Of course, if you are experiencing a very tight schedule, it should be totally valid to use and evolve a single class diagram from start to end, and just convert your analysis classes into tables, and delete collapsed or unused classes from your diagram: just keep the analysis and implementation versions of your EA project under different names.
On how to relate your analysis classes to implementation classes, see:
http://www.sparxsystems.com.au/cgi-bin/yabb/YaBB.cgi?board=general&action=display&num=1034609727I hope this helps.
Jaime Gonzalez