Why not specifying that arguments in a procedure call should evaluate to exactly one value ?
Similarly binding init would evaluate to only one value.
The R5RS (and current draft) make it very specific that passing MV to a continuation other than that created by call-with-values is an error. Unless you want to strengthen this and require it to signal an error (thus making extensions like the CL MV behavior) then there's nothing to vote on.
No, the problem is about allowing or not values to interfere with another one. For instance :
((lambda (X Y) ...) (values A B) (values))Do we allow implementations to bind A to X and B to Y ?
WG1 rejected this proposal.
IMO, better to take the R6RS language that the continuations of the arguments to a procedure call expect a single value, and that passing multiple values to a single-value continuation context is unspecified.