Presented in this paper are algorithms for software agents who buy components and sell services through auctions. The problem is important when an agent has to make decisions about what to buy at what price, prior to knowing if what was bought can be combined with other components to be resold for a profit.

The paper presents two algorithms--one more exact and the other more efficient--based on computing the expected utility from the auctions. The use of the algorithms by an imaginary transport company is described. The paper is clearly written, and provides sufficient details about the algorithms that would enable others to implement them. The assumptions made in the study are also clearly stated. It would have been nice to have more experimental results, as well as results on real data, to be able to assess how well the agent using the algorithms performs, but it is hard to expect more from a conference paper.