EDI, XML, and the Transparency Problem in Electronic Commerce

Standard (that is, long-standing and currently much in use) EDI protocols (including the X12 and EDIFACT series) have repeatedly been criticized for poor design, confusing or absent semantics, and much else. Most of these criticisms are indeed on the mark. The main conclusions that the critics have drawn are also correct: business-to-business e-commerce is expensive and difficult to set up and maintain, because of shortcomings in the design concepts underlying standard EDI. Something must be done, but what? Central to the problem is the fundamental question of semantic transparency: When A sends B a message, how does B’s machine know what the message is about, what it means? Given proper standards, message meanings are determined and computers can be programmed to act appropriately to the intended message meanings. The complaint against EDI has been that proper standards cannot be made because of the misguided way in which the EDI standards are designed. Proponents of XML have been touting XML’s strengths and claiming that they overcome, or can overcome, the semantic transparency problem in e-commerce. In support of this claim, proponents point to the DTDs (or similar devices) that any XML/EDI solution would use. The claim is that semantic transparency is/can be achieved through the DTDs. In this paper I argue that indeed the DTD mechanism offers a kind of progress on the semantic transparency problem, but that it cannot provide anything approaching a complete solution. While XML+DTDs is indeed a very promising vehicle for structuring and transporting messages for business-to-business commerce, it is not itself a semantic theory of what those messages say. We need the semantic theory. Once we have that, XML can be used to embody it for applications. Drawing on previous work, I will present the elements of my formal semantic theory for business messaging (the “lean events theory”). With examples from this theory before us, we can get a more proper view of the semantic transparency problem (aka: the spanning problem). This is not a problem that can be made to go away entirely, but we can live with it and do commerce.