[Haskell-cafe] Derived Read for records

Someone at work just asked me about the inflexibility of the derived Read instances for records -- specifically that they require fields to be given in the same order as in the type definition and that fields cannot be omitted. I hadn't been aware of these restrictions. A few questions: * Are there known work-arounds? * Is it particularly difficult to synthesize more flexible Read instances? * Do people just not mind the restrictions? Thanks, -- Conal