Iceberg transport costs are one of the main ingredients of any modern quantitative trade or economic geography model. Making the iceberg assumption: to deliver q goods produced in location i to location j, τijx goods have to be shipped from location i (τij > 1), one avoids the need to explicitly model a transport sector. In this paper we assess some of the assumptions and implications underlying iceberg transport costs using data from the only business where the product literally melts in transit: the long distance trade in natural ice. The ice trade flourished in 19th century Boston. From its harbor large quantities of ice were shipped over vast distances, even to tropical locations in Asia, South America and Australia. Using detailed information on this business from notably the company records of the largest Boston Ice company we verify the extent to which the iceberg assumption provides a good approximation to the actual cost of shipping icebergs.