Good discussion. Let me give some insight into organization and recommendations.
Here I'll just address numbers of "data elements." I'll write up a formal "data structure" recommendation for discussion.
For any given state we have 300 units formed generally - 3000 for each of the the Union and Confederate sides. The number of units who formed across state lines is minimal - about 2% on each side. Each unit was made up of Regiments or batteries in the case of artillery and each regiment consisted of 10 companies generally. Each regiment was assigned to a Brigadier General in groups of 3 to 5 and brigades were assigned to divisions and armies. Each company was generally formed within a particular county of which we had 2,000 at the time of the Civil War and this is the organization level for identifiying family units as you will generally find all family members serving in a unit formed from their county. I have identified unit designations by county on my web site for Georgia, South Carolina, and Alabama. This has generally not been done for the other states.
There were 20,000 events (battles, skirmishes, etc.) in the war which can be organized by state, by date, by opposing armies or by campaigns.
Finally we have 3 million men on both sides who fought.
So to properly map out this project we need.
A state heirarcy showing units formed, county designations, and battles fought within the state, and bibliography of unit histories, diaries and journals, battles, county histories and genealogies.
A battles heirarcy showing "orders of battle", casualty lists, cemetery listings, bibliography, and first hand accounts .
An army heirarcy linking to the brigadier generals, their commands, their orders of battle, battle accounts and bibliography for research by this army structure.
A smattering of each of these data elements are scattered around the internet.
The FIRST question to be answered is are we going to create another "smattering" here or are we going to tackle the big picture. I don't mind contributing my research which currently approaches 20,000 pages on line, but so far I have found the Geni interface a bit tedious to work with and more than I want to tackle on my own.
John Rigdon
www.researchonline.net