Yes. I would still use a RDBMS.
I would design the application *FOR change*.
By this I mean: I will try my best to model as many concrete attributes as
possible at the beginning. Where the attributes start growing into
hundreds - I'll hang them off as 1:1 tables off the "main" table. So, in
your example: Person table. Then have separate table for Favourites,
Hobbies etc. Same key - different specific attributes. If Favourites is
broken up into colour, foods etc - go one more level down with the 1:1
mapping. This I will try to do UP FRONT - as much as possible.
BUT!!! - i will also definitely write a special program to cater for the
"maturing" attributes so that every once in a while, I would
programmatically generate the necessary tables and do the data migration
into fully-fledged tables and first class attributes. Your application
should be dynamic enough to switch to and from the two different types of
data structures without reprogramming. You need only write this
functionality/architecture once. Use your own naming standards etc, stick
to them, identify patterns (force them sometimes) and automate the process.
It works.
hth
Regards
Abdullah
