My OOPSLA CD-ROM is at home, but I recall Henry Lieberman had a paper in the 1986 conference illustrating what can be done using prototypes that cannot be done as well using classes. If I recall correctly again, this led to debates and ultimatelyf (was it OOPSLA 1987?) something called the Treaty or Orlando (the site of OOPSLA that year) which lays out the relationship between classes and prototypes, at least at that time.
Technically, the issue wasn't classes vs. prototypes, but it was inheritance vs. delegation. But it hits the same mark, eh?