Lassi on loss of precision: as we noted offline, it is often important that two timestamps be precisely the same, which is not reliably possible if they are passed through a float representation.

Marc's numbered comments:

1) I have added language allowing identifiers to be bound to different (more strictly conforming) procedures and macros than the implementation itself provides.

2) read and write are gone

3) I think a Scheme that coalesces quoted or self-evaluating input and yet allows mutation would not be a conforming Scheme in the first place, and I do not know of any such Scheme, so I'm not going to add language for this.