Hmm, it sounds like you're trying to argue with me, but I'm not clear on what you're saying — some quick responses — hope this helps:
① I did not claim Scheme was an "assembly language" for ML. I said that programming in, reading programs in, and compiling programs in Scheme was like programming in, reading programs in, and compiling programs in assembly language in some specific ways, and very unlike it in others. The relationship between Scheme and ML is that some of the central insights of Scheme were adopted by ML.
② ML defines evaluation order. The lambda calculus does not. Typical ML implementations compile to the assembly languages of actual processors. Can you clarify?
③ I think Scheme is sufficiently well defined for this discussion — it's a family of languages originating in some papers by Sussman and Steele in the 1970s, and continuing through the current R7RS work, including a number of compilers. Several of the standards define the semantics of the language symbolically, not just in English.
① I did not claim Scheme was an "assembly language" for ML. I said that programming in, reading programs in, and compiling programs in Scheme was like programming in, reading programs in, and compiling programs in assembly language in some specific ways, and very unlike it in others. The relationship between Scheme and ML is that some of the central insights of Scheme were adopted by ML.
② ML defines evaluation order. The lambda calculus does not. Typical ML implementations compile to the assembly languages of actual processors. Can you clarify?
③ I think Scheme is sufficiently well defined for this discussion — it's a family of languages originating in some papers by Sussman and Steele in the 1970s, and continuing through the current R7RS work, including a number of compilers. Several of the standards define the semantics of the language symbolically, not just in English.