The ontological representation of products and services is a core challenge on the road to business applications for the Semantic Web. This will not only help search engines provide more precise product search for human users, but can be expected to support a much higher degree of business process automation in general, especially in all tasks that involve content integration. In industrial data interchange between business partners, the state of the art is the use of common XML schema definitions (e.g. BMEcat) for the representation of structure and the use of classification schemes (e.g. UNSPSC or eCl@ss) for the representation of product semantics. This current practice, however, takes place in well-defined contexts known to both the publisher of data and the recipient, which allows even the usage of the same standard with varying semantics in distinct settings. In a Semantic Web context in contrast, the same document must be machine-readable (1) by a huge number of different partners (2) for a multiplicity of purposes. In other words, the data recipient and the data usage are not predetermined, which makes it much more difficult to reach consensus e.g. about suitable product classes. This paper develops the requirements for product representation in the Semantic Web and evaluates existing alternatives.