Saying I miss the whole idea about Perl is a rather strong thing to say.
I didn't say the linguistic design was a happy accident after Perl was designed--what I did mean to say is that it was clever of Larry to start with the linguistic approach and derive the language from that. Which is basically what you said, so we agree. :)
Although, however nice these linguistic features can be in the small, I do think the linguistic approach ultimately breaks down since natural languages cannot be represented by a Turing machine. What we can represent on hardware as being computable is an infinitesimal subset of what we can represent in our minds with non-context-free(and context-sensitive) grammars, so I would argue that the real benefits and optimizations of the linguistic approach cannot even be realized in a programming language. (That is until some genius invents a non-recursive hyptercomputer
.) And please don't take this as a hard opinion--I do recognize it's a fuzzy field in which I have no expertise, just interest.