Re: why hierarchy?

Neo wrote:
> > I suspect the former isn't possible as it would necessarily require some unknown to become known for any demonstration to work. This would seem a fallacy if one agrees that the known can't also be unknown.
>
> T1 (ie today)
> Have someone specify a relatively simple problem (ie model john and
> mary). You post SQL script to store/query data. I post dbd's script to
> store/query data.
>
> T2 (ie later today or tomorrow)
> Have someone specify additional data requirements. You/I show/compare
> the impact on data/script to accomodate new requirements.
>
> The above simulates a customer specifying new/additional requirements
> after project has been updated to meet customer's prior requirements.

Let's start simple. Let's see how the relation model (using either a pure relational syntax or a proprietary SQL implementation) and "dbd" model work against some simple requirements. We'll first evalute the two based on how well they handle the initial specifications.

Basic Requirements: A business has many staff members (with full names and a start date), some of whom are direct supervisors over other staff members. The business has many clients (with full names), and each client is registered on a specific date by a specific staff member.

Constraints: Assuming a few simple forms exists for directly entering new data into the system, the following constraints should be enforced w/out embedding the logic into the forms: It should not be possible for a staff member to have multiple direct supervisors.
A client cannot be registered twice.
A staff member cannot register a client before that staff member's start date.

Queries: It should be possible to supply queries which resolve the following questions:
'Alice' is retiring. List all of the clients she signed, and all clients signed by employees whom she directly or indirectly supervised. We need to make some staffing cuts. List all current employees who have signed less than 10 clients in the past month, in order of seniority.

Let's first evaluate how well the two models hold up relative to the intial requirements. Obviously, if one model is markedly inferior with respect to initial requirements, then who cares how well it responds to changes in requirements? Received on Thu Jul 27 2006 - 15:21:47 CEST