Perhaps we need to add a "rule" element as Bob Morris has suggested. Then with that additional fact, the usage of the other terms would be specifically declared by the provider and all this assumption/inferencing would not be needed, where the declaration of the rule was provided.
I don't follow. From a DwC perspective, the providers are serving text strings. I think the proposal by David and Markus adequately captures the needs of both the providers, and the consumers. It's a good compromise solution (80% of the benefit with 20% of the work).
But, millions of rows of legacy data may never conform to anything done at this point. If the meaning of ScientificName is altered by a definitional change after 10 years of the DarwinCore term being used with a different definition, no doubt the end result will be even more world-wide data hegemony because there will not be a sudden switchover of all the legacy data to the new definition. That herd of elephants is not going to turn quickly, so for some long time you really won't know what you have in a given ScientificName field - the old definition or the new.
That was the basis for my original hesitation to redefine scientificName; but here's the thing -- over those ten years, the term has *not* been consistently applied or used. The herd of elephants has just been meandering aimlessly in this sense. What I think the proposed solution allows is at least a start of orienting the elephants heading in the same general direction. The point is, as evidenced by the data GBIF harvests, we *already* don't know what we have in a given scientificName field, because providers are not applying the stated definition consistently.
Aloha, Rich