Recently, several languages for web service composition have emerged (e.g., BPEL4WS and WSCI). The goal of these languages is to glue web services together in a process-oriented way. For this purpose, these languages typically borrow concepts from workflow management systems and embed these concepts in the so-called "web services stack". Up to now, little or no effort has been dedicated to systematically evaluate the capabilities and limitations of these languages. BPEL4WS for example is said to combine the best of other standards for web service composition such as WSFL (IBM) and XLANG (Microsoft), and allows for a mixture of block structured and graph structured process models. However, aspects such as the expressiveness, adequacy, orthogonality, and formal characterization of BPEL4WS (e.g. reachability) have not yet been systematically investigated. Although BPEL4WS is not a bad proposal, it is remarkable how much attention it receives while more fundamental issues such as semantics, expressiveness, and adequacy do not get the attention they deserve. Therefore, we advocate the use of more rigorous approaches to critically evaluate the so-called standards for web services composition, and to learn from 25 years of experiences in the workflow/office automation domain.