Database Design Article?

Not really answer to your question. I think I saw in iSeries Network. 

The concept is what I call field select logicals and the reason for
doing it is data base independence. It, also, gives you better I/O
performance. 

I used it quite a bit on building systems before SQL really took off. 

The disadvantage to doing it is that you have to create one logical for
each program and each file or join files. Ton of work. About the only
way to make it practical would be to create a tool that allowed you pick
files and fields and generate the logicals automatically.  

All you are really doing is giving the record format name a unique name
and then entering the names of the fields that you want included. 

     A                                      UNIQUE

     A          R RIDAT                     PFILE(XVSTTENT)

     A            TETABLEID

     A          K TETABLEID


The much better solution, use SQL and bring only what you need into the
program and you have your database independence and potentially better
performance. 


<snip>
Essentially, the author was advocating using logical files with only the
fields the application is using.  This would enable you to add
additional
fields to the physical file without the need to recompile every logical
and associated programs.
</snip>