[
Lists Home |
Date Index |
Thread Index
]
Bob Foster wrote:
>
>But relationships do change all the time. <#Simon> :sonOf_:1 is as good an
>example as any. I assume the 1 is intended to be the cardinality, so the
>relationship might be better named sonOfFather.
>
No, but no matter. It just says that "Simon is the son of someone."
>But if your mother divorces
>and remarries, sonOfFather may no longer be many-to-one, but many-to-many.
>The latter probably better captures the relationship in the US, where
>roughly half of first marriages end in divorce.
>
A few thoughts on this train of thought come to mind:
1) the 80/20 rule applies to everything i.e. even though there may be
edge cases that invalidate some particular classification, nonethess the
classification might be useful.
2) When someone's parent's divorce, the "father of" relationship remains.
>Wasn't so in the early part
>of the last century; it's a culturally induced change. Moreover, if you are
>cloned, you may not have a father, at all; which is quite a different thing
>than having a father but not knowing who he is.
>
3) hmmm... I haden't considered who the father of a clone might be...
that falls into the 99.99999/0.00000001 rule.
>Now consider the "type" of the father, nominally a male Person. Suppose a
>woman is impregnated by a sperm donor at a fertility clinic, where the
>identity of the donor is kept secret. This could be modeled as an unknown
>father, but in fact the identity of the biological father is known by the
>clinic, and the name of the clinic can serve as a surrogate for the father.
>If the type of the father is changed to the union of Person|Clinic, the
>parentage information is not lost. More children than one might suppose
>would benefit from this change.
>
>
4) One might distinguish between "legal father of" and "biological
father of". I've already provided for the fact that the "father of"
someone might be unknown. Indeed I've never met Simon's father, and have
no knowledge of *any* details about him save for the fact that he
fathered Simon.
>The point about relationships being neither homogeneous nor stable over
>time, in either number or type, nor universally applicable, has been well
>made by William Kent, e.g., in his classic "Limitations of Record-Based
>Information Models" ACM TODS paper.
>
5) Nonetheless, someone else once demonstrated that classifications are
useful.
Jonathan
|
