Identification of COTS candidates is a complex activity itself. It implies not only dealing with an impressive number of possible candidates but also with unstructured information that requires a careful analysis. In this context, some proposals currently use description logics to develop an ontology for matching requested and provided services, or propose taxonomies for classification; others suggest extending the identification stage with a learning phase, which provides support to the COTS component discovery process. In this paper, we present common features and differences among some of these proposals, identifying relevant issues for COTS component classification and identification in the near future.