Two substantive changes were made between the ninth draft and the tenth (post-ballot) draft:
- The R5RS semantics of eqv? as applied to procedures was restored: it must report true if they have the same location tags, and must report false if they have different behavior (results or side effects) on some arguments, but may report either true or false when comparing other procedures
- The eq? predicate is now required to report true when two procedures have the same location tag, but may report either true or false when comparing other procedures
These are the editorial corrections to the ninth draft in reverse order. All are present in the tenth draft.
- Covered additional language changes in the changes list
- Removed s, f, d, and l from the formal syntax of an exponent marker
- Added #true and #false to language changes
- Corrected commenters' names
- Changed rationale of eqv? to be normative
- Improved language that defines when inexact numbers are not eqv?
- Cleaned up bibliography to include only cited works, including SRFIs
- Added grammar for case-lambda to tail-recursion section
- Warning about <init>s was incorrectly on letrec instead of letrec*
- Fixed minor editorial nits from a last-minute review
- Updated acknowledgements of commenters
- Rule out xbeef as a character name because it is a hex scalar value
- Restored R5RS results of gen-counter and gen-loser
- Cleaned up definition of "variable definition"
- Cleaned up zero-or-more values continuations language
- Falling off a cond-expand is now undefined behavior (no particular behavior was voted on)
- Minor typos fixed (thanks to Jussi Piitulainen)
- Corrected changes not to imply that character and string comparisons accept 0 or 1 arguments
- Added note about the order of arguments in R6RS vs. R7RS bytevector-copy!
- Restored reference to eqv? in the storage model (as opposed to "operational equivalence")
- Improved allocation discussion to handle empty objects specially
- Changed "as following" to "as follows"
- Cleaned up all cases of "forbidden in strings" to be uniform, removing redundant references
- Made the definitions of character comparisons defer to char->integer to handle non-Unicode
- Fixed quasiquotation and symbol-equality examples
- Removed procedure inequivalence from language changes list
- Squeezed vertical space to save a whole page
- Removed "equal," from definition of string<? and friends