Here is a complete discussion on the character set of cookie name (for python development):
http://bugs.python.org/issue2193

I think drakma should follow python's work.

In file "util.lisp" of drakma, function split-set-cookie-string splits cookie string using function is-valid-cookiename-char which comes from chunga.
But function is-valid-cookiename-char just follow RFC2616 which rules that chars of cookie name are legal when they are in token set ( see bellow).

  token          = 1*<any CHAR except CTLs or separators>
  separators     = "(" | ")" | "<" | ">" | "@"
                      | "," | ";" | ":" | "\" | <">
                      | "/" | "[" | "]" | "?" | "="
                      | "{" | "}" | SP | HT

That is not practical!

In python, letters, digits and "!#$%&'*+-.^_`|~:" are legal in Cookie name.