But there is a problem. The relative importance of the individual terms in facet A, expressed in hierarchy or sequence in array, may be quite different in another facet. Leaving chemistry to look at another domain, a “domestic animals” facet would not want to use the details of a zoology facet A consisting of a standard taxonomy. In that case, a separate listing of terms must be constructed for the “domestic animals” facet (Ranganathan used to call these terms “favoured categories”). We may add to this listing the term “Other—divide like facet A”. This is equivalent to listing the terms in the facet in the order of their “importance” in the subject field, while allowing extra terms to be drawn in for animals that are occasionally domesticated.
(f)
One issue that has been discussed by Wilson (2006) may be called, following his example, the “chocolate-pecan problem”. In a classification of confectionery, he lists a facet “flavours” such as cherry, chocolate, pecan. How is the flavour mixture “chocolate and pecan” to be handled? In a classified catalogue, the mixed flavour could be represented by linking the notations for the two flavours (e.g. cho-pec). This cannot be done on the Web, and Wilson rightly argues that to assign both terms to an entity at the classifying stage violates the principle that the terms in a facet should be mutually exclusive. Such an assignation at the search stage is better interpreted as a call for “chocolate OR pecan”.
One solution might be to include mixtures such as “chocolate/pecan” as terms in the flavours facet. I do not consider that this does violence to exclusivity—a mixture is a different flavour from its constituents, or why bother to provide it? Would we reject the term “sweet and sour” in Chinese cookery? An alternative solution is to add a term “mixed” in the flavours facet, and provide as a “subfacet” a checklist of the original flavours with the instruction “tick each flavour in the mixture”. Any combination could thus be constructed and used in classifying and search—though admittedly this would require programming beyond the basic facet manipulation.