(add-to-list 'process-coding-system-alist '("zsh" . utf-8))

[kj <no.email@please.post>]

I want to set once and for all (in my .emacs, that is) the coding
system for all the "inferior shells" I start with M-x shell.

As the subject line above says, I tried

  (add-to-list 'process-coding-system-alist '("zsh" . utf-8))

(my default shell is zsh, of course), but it did not work: new
shells still mis-handle non-ASCII characters until I put them
through the usual C-x RET p business.

I would like to know why the elisp snippet above fails to make the
initial coding system of new shells created by M-x shell to be
utf-8.

Any ideas?

(NB: there may be other ways to achieve the same result, but here
I am specifically interested in understanding why what I tried did
not work, and fix it.)

Thanks!