Dan Bornstein:
> This seems like a good enough reason to "use up" another character to me:
> It's a new semantic distinction, and it is going to be used fairly
> frequently in SRFI-1 and, no doubt, other SRFIs too. Part of being a "good
> saver" is knowing when the right time to spend is, and to spend it on good
> things.
>
> I vote for "@".

I don't believe we are dealing with new semantic distinction here: "linear
update" is just another case of *underspecification*, not a completely
new semantics. The idea here is not to specify a new way of handling
lists but to discourage users from relying on certain implementation
details. I believe that most implementations will indeed use destructive
update implementations, and I don't thing there is a point in specifying
how exactly those cons cells are modified.

I vote for bang!