The main difference with Napster is that, in the case of music, there was some argument to be made that piracy was harmful to music, because some (tiny) part of music sales benefited the artists (and, as we all know, every pirate download was a lost sale...).
In the case of science publishers, the money simply never goes from publishers to researchers. Scientists are paid by entirely different actors (who ironically also end up paying publishers, for subscription fees to access papers). So the only possible argument left is that publishers are somehow beneficial to the science publishing process, which is a much weaker argument.
Imagine if music sales did not bring any money to artists at all, but the music industry argued: "sure, music would get composed and recorded without us, but how could it be distributed without us?" This is the situation science is currently into. The only explanation to the current state of affairs is history (pre-Internet science distribution was complicated) and the unimaginable inertia of prestige.