Re: Modelling and the element of time

"Drago Ganic" <drago.ganic_at_in2.hr> wrote in message news:9op6o6$3iud$1_at_as201.hinet.hr... Hi there,

I'm modeling a general business system (party, product, order, shipment, invoice, payment etc.) that will be sold as a product, but parts of it will also be used in other products/projects my company sells. So, I try to solve some general and known problems that can be used by anybody in my company.

The system will contain two parts: a production system (mainly for data entry) and a data warehousing/data mart system (only for reporting and analysis).

I would like to hear some opinions or get some references about modeling the element of time. That problem is solved very nice in the DW concept. But I have also a need for it in the production system.

There is also another problem - that is connected with the time problem - Should I model the complexity of the real world ??!?

:

hi drago -- greetings to croatia

several years ago a colleague and i were engaged to write an hr tracking and query system which would model employees in an organization. the basic intuition was this: an individual in the firm possesses various properties which vary over time. for example, a particular individual occupies various offices at different moments in the course of his employment. we chose to represent this as an assertion of the form 'P(i) from f-time to t-time'. that is, as a state holding over an interval. events, which occur instantaneously, can be represented as assertions of the form 'P(i) from x to x', where f-time = t-time.

moreover, such assertions are themselves dated. for example, on january 3 1999 we assert of tom that he occupies office 17 from october 4 1998 through some unknown future date. when, on february 13 2000 tom moves to a new office, we record a further assertion to that effect.

finally, we may occasionally need to add new properties to the system. for example, the firm adopts a new savings program in which individuals may participate.

the collection of all assertions of the form 'on a-time P(i) from f-time to t-time' constitutes the raw data. notice that you never delete records. later assertions mask (wholly or partially) earlier ones.

since the time of our work on this project a considerable amount of material on the subject has appeared (i have a copy of snodgrass &al's _temporal databases_ on my bookshelf, but it was published too late to be of use to us.) i wouldn't be surprised to learn that "bitemporal"  extensions to commercial db systems have since arrived.

drop me a note if you'd like to get a copy of our whitepaper describing the design in more detail. Received on Tue Sep 25 2001 - 15:02:12 CEST