Patents constitute our foremost policy tool for encouraging innovation.However, because each new technology provides an important input tosubsequent innovation, the exclusive rights conferred by a patent may alsoimpose significant costs upon follow-on innovators. Optimal patent policyshould seek to maximize the patent incentive effect, while minimizingburdens placed on future innovation by tailoring the scope of the patent tothe characteristics of each technological sector affected.In the case of software, recent scholarship has illuminated the innovationprofile of the current industry. Software is characterized by incrementalinnovation, relatively low development costs, and short, volatile productlife cycles. Interoperability and compatibility between complementaryproducts is a major concern, making technical transparency or reverseengineering critical to product development. This suggests a need forrelatively narrow patents that are relatively easy to obtain, and subjectto the exceptions necessary to ensure interoperation and follow-ondevelopment.However, current software patent doctrine bears little relationship to thisindustrial profile. The United States Court of Appeals for the FederalCircuit has set an extremely lax standard of disclosure software patents,resulting in patents scope unconstrained by doctrines of enablement andwritten description. Recent changes that make patent law amenable tosoftware have produced a flood of new applications, allowing firms to adopta patent thicket strategy for licensing leverage. At the same time, FederalCircuit case law suggests that a stringent standard for patentnon-obviousness will be applied to such patents, resulting in relativelyfew valid software patents. Optimal software patent doctrine wouldconstrain scope to deal with patent thicket while lowering thenon-obviousness standard to validate more issued software patents.