(RQT) Character Dependency rules

[Gregor Hagedorn]

> DELTA already allows dependencies between characters - such
> that if a specific state has been selected for a character, other
> characters are ignored (i.e. no wings, ignore anything to do
> with wings). Is there any other dependency relationships that
> might be required, or additional information about such a
> relationship?

Leigh brought this up; I would like to start a new thread with this.
What problems do we have with dependency relationships (or integrity
rules, to put it another way)?

Most of us are familiar with the dependency as implemented in DELTA
compatible applications. here one character is defined as
controlling, and that depending on which states is scored in a given
item, other characters become applicable or inapplicable in that
item. Questions:

1. Do we need both applicable and inapplicable definitions?

In DeltaAccess I assumed that applicable is the complement of
inapplicable. If the complement of applicable states are defined as
inapplicable, the application behaves the same way. This is only
partly true.

a) If new states are added in the definition, some care is necessary.
I realized that in DeltaAccess, but thought it the lesser problem.
However, for the same reason for which we may need actually a "not"
statement in the item description (a statement "color of X is NOT
pure white" remains true, regardless of how many shades of beige are
later added later on)

b) If multiple state are scored, the behavior needs to be defined. In
the CSIRO programs, it seems that if two states are scored in a given
item, a single one is sufficient to make another characters
applicable or inapplicable (which makes sense, but is lost if the
complement is choosen).

I have not found a good generalization for this, but loath to
implement the rules in duplicate. Any ideas? This is certainly a
question I am thinking about on the logical rather than conceptual
level, but answering it may help to define the requirements, i.e.
whether the DELTA applicable/inapplicable model should be followed,
or whether alternative expressions are possible.


2. I recently had a good discussion with Wouter Addink and Flip Boer
in Holland. Among other things we considered whether a structural
hierarchy (leaf stalk is part of leaf is part of plant) could replace
dependency rules. Clearly, it can replace some rules: if there is no
leaf, all leaf characters will be inapplicable. However, it can not
replace depencies based on multiple states that are not
present/absence. Further, it may not always be true because we tend
to use confusing terminology: Something at the leaf stalk base would
belong to leaf stalk, but absence of leaf stalk (i.e. absence of
measurable lenght of leaf stalk) does not imply absence of stalk
based characters. Any further ideas on this?


3. (and finally): Does anybody have good examples where character
_state_ dependencies would be necessary? This has been proposed for
New Delta. I have no real objections, except that I have not found a
good example where I would want to have it.
----------------------------------------------------------
Gregor Hagedorn                 G.Hagedorn at bba.de
Institute for Plant Virology, Microbiology, and Biosafety
Federal Research Center for Agriculture and Forestry (BBA)
Koenigin-Luise-Str. 19          Tel: +49-30-8304-2220
14195 Berlin, Germany           Fax: +49-30-8304-2203

Often wrong but never in doubt!