Design of Temporal Databases

Greetings:

        I have some questions concerning the design of temporal databases (tdb) and any help would be much appreciated.

        I have a temporal metadata system that allows for modifications w.r.t. time. Basically, the key composition for some parent table is as follows:

        code + start_month + end_month

        Thus, specific codes will be "live" for specific periods of time, and any derived children will themselves be "live".

        My questions are:

1. What are the temporal rules, especially when the time frame of the parent can be altered? Recall, all derived children must still remain totally live for the new time frame. Our initial design places the time element within the primary key. Any issues with this? (such as performance, etc) Another design involves creating a time table, with various code tables now dependant on this time table. In otherwords, the time table is normalized. What are the pros/cons when compared to having a key composed of time?
2. Can anyone see an advantage w.r.t. the actual key composition (assuming we go this way):

        key: code + start_month + end_month
                        vs
        key: start_month + end_month + code
                        vs
             other permutations   

c) Which data type should we use for the start/end month (century and month

   only; no need for day and/or time). Our dbms is Oracle 7.0, and possible    types include:

        Date (Oracle object)
        Number 6
        Char (6)

Any help and/or comments would be much appreciated.

Thanx in advance,

gerry
.

-- 
gerald c. hurdle (hey, it's a name...)
internet: ar443_at_freenet.carleton.ca
phone: (613) 820-1873
philosophy: nothing is true!