Re: The original version
Date: Mon, 27 Dec 2010 16:55:16 -0800 (PST)
Message-ID: <2e005595-a617-4fb3-b298-66f8af928d08_at_o14g2000yqe.googlegroups.com>
(i)
When we talk about modeling relationships in Anchor Modeling, then the
following question arises: what do the authors of Anchor Modeling
model?
If we have a relationship between two entities, then we should notice
that both entities change over time. Each entity can have for instance
10 attributes and each attribute changes independent of time. Also,
the relationship itself can change with time. In my paper I do not
model relationships between entities, but relationships between the
corresponding states of two entities (or relationships).
In their basic version from 2009 Anchor Modeling defines relationships
in the following way:
Def 9 (Historized Tie). A historized tie Htie(C1,…,Cn, T) relating a
set of
anchors {A1(C1),…,Am(Cm)} is a table with n + 1 columns satisfying n
>= m
and n >= 2, where for every i in [1, n], Htie.Ci is a non-null foreign
key to some
Aj.Cj for j in [1,m], and the domain of the last column T is a non-
null time
type. The primary key for Htie is a subset of (C1,…,Cn, T) containing
T.
(ii)
According to Leibniz’s Law, every entity becomes another entity when
its value changes.
Does Anchor Modeling work on an audit file level or on a database
level?
(iii)
It is possible for two men to possess different values for a certain
property, for the same period of time. Similarly, this can happen for
a relationship. A manager wishes to record both of these values. He
wants to maintain statistic of bad workers. He also wants online db.
The following case is for instance possible in court:
One side claims the car in an accident was blue, the other claims it
was red, while the police claims the car was black. The court requests
that all of these three colors be recorded in the database. Obviously,
Anchor Modeling cannot solve this.
Vladimir Odrljin Received on Tue Dec 28 2010 - 01:55:16 CET