Margherita,

this is good approach when possible.

Because of the notational constraints - older schemes will have the following two situations/problems that can only be resolved through IDs - history of concept position in the scheme and history of labels - I give examples of UDC but this would be similar to all classification schemes that maintained for more than 50-100 years and are going through necessary restructuring to follow the development of science:

1. change of label (depricated labels)

In 1905 concept ID*00001 introduced represented with label A

Upon restructuring of the scheme in 1975 concept ID*00001" is moved to a new class (different hierarchy) hence this concept gets a new id ID*00999 and it gets a new label X (i.e. label X replaces label A (depricated) - X label is new (never used before) Relationship between ID*00001 and ID*00999 has to be established - but also the relationship between labels A - X

2. re-use of once depricted labels to denote new concept

In 1990 new concept ID*01333 introduced - and label A is re-used for this new concept (label A was before used to represent concept ID*00001 which is now represented with label X and has ID*00999)

In practice and in real life there will be collections using A to represent concept 00001 (old classification schedules) and those using A to represent concept 01333 (new classification schedules). Three will be need in disambiguating A(s) and linking IDs.

I now realise that two IDs for concept that *is* and *is not* the same :-) is also a very good example of what Johan calls 'concept versioning' - I would only say that it is better to use expression 'scheme versioning'

I am aware that the management of label version may be simplified and made more logical and elegant - but this would be a luxury only modern schemes can afford. Old data and its use pose many constraints.

Aida