Taxonomic hierarchy in SDD

Eric Zurcher Eric.Zurcher at PI.CSIRO.AU
Wed Nov 28 09:54:06 CET 2001

One additional point to consider is that hierarchies arise not only in the
classification of items (taxa or specimens), but also in the classification
of the characters used to describe those item. And along both the item and
character "axes", there will often be multiple, alternative hierarchies
which reflect different purposes or philosophies.

For example, in assembling an interactive key to plants, one might well
want to group characters dealing with leaves separately from characters
dealing with the infloresence. And perhaps a trickier "hierarchy" of
characters is one that might be used to describe the rules for "natural
language" generation, aggregating characters into phrases, sentences and

Does anyone see a good generalized mechanism for handling the creation and
maintenance of hierarchies (or other groupings)?

Here are a few thoughts off the top of my head: Because of the need for
handling alternative hierarchies for different purposes, it would seem
logical to keep the "classification" separate from the "core" data (going
back to DELTA, this was one reason why "directives" files were used - it
provided a mechanism for changing groupings without touching the core
data). But this approach comes at a significant cost: if the "core" data is
changed, it will probably require synchronized changes in the various
hierarchy descriptions which reference that data. Maintaining consistency
becomes a problem. Perhaps the way out is to make use of unique global
identifiers (GUIDs), as Steve Shattuck has suggested.

At 10:43 PM 26-11-01 -0500, Bob Morris wrote:
>We may not be arguing here. I don't dispute the need for this, only
>argue that there should not be a standard for taxonomy, but rather a
>standard for how to specify taxonomy. Or any other hierarchy for that
>Doing this would allow you to get the hiearchy from /any/ suitable
>source, including but not limited to the data source itself. To me it
>seems that the least brittle thing for data inheritance is to use the
>same model as for datatype inheritance, i.e. a separate "schema" to
>which reference is made, as Tim Jones argued.

Eric Zurcher
CSIRO Livestock Industries
Canberra, ACT Australia
E-mail: Eric.Zurcher at

More information about the tdwg-content mailing list