The SPICE project is developing a globally distributed federated database of biological knowledge, forming a ‘catalogue of life’ by harnessing specialist expertise on classification of groups of organisms. The component databases are heterogeneous, and are joined to the federation in various ways. We explain how our federated approach partitions the task of maintaining a consistent classification into manageable sub-tasks. We use both CORBA and XML and, while CORBA is widely used for interoperable systems and XML is attractive for data exchange, some problems have arisen in practice. We discuss the problems encountered when incorporating CORBA ORBs from multiple vendors, compromising true platform independence. We also discuss the non-trivial effort required to achieve stability in CORBA-based systems, despite the benefits offered by CORBA in this respect. We present preliminary results, illustrating how performance is affected by various implementational choices.