E-commerce, especially information commerce, allows a consumer to purchase "bundles" of goods from many different, autonomous suppliers.  In some cases, the customer is buying from brokers and does not know the ultimate provider of the goods.  In an environment of distrust, these "distributed commerce transactions" must be structured to protect the interests of the customer (atomicity of purchase) and brokers (no buyers backing out).  We present a formalism for describing these transaction problems, and a distributed algorithm for providing "safe" solutions where they exist.

     Unfortunately, there are many cases where completely safe solutions do not exist.  Therefore, we relax the guarantees provided by the algorithm, looking at expected utility rather than worst case.  This model permits the selection among suppliers that differ in price and reliability.
"Request strategies" guide the customer in placing orders to maximize expected utility.  We present different representations based on whether the customer faces a deadline, and whether there are multiple goods, and multiple suppliers per good.  We also present heuristics to reduce the size of the strategy space when searching for an optimal strategy.

(Joint work with Prof. Hector Garcia-Molina)