> At the moment, SRFI 219's `define` form is defined through its
> implementation. It would be better if, in the final version, `define`
> was specified abstractly (much like the special forms in chapter 4 and 5
> of the R7RS), so that the implementation could actually be tested
> against the specification.

Good idea. I can do that later; I hesitate to send three drafts in 24
hours :)