bug#19226: eww.el desktop support fixes: autoload eww-mode, use inhibit-read-only

[Stefan Monnier]

>       Yes.  Yet I fail to see how that would be an improvement?

Maybe it's not worth the trouble, indeed.

>       changes.  Now, we also make that fail when it’s the /name/ of
>       the handler function that changes.  (I presume that the change
>       in the mode name is much less likely on one side, and much more
>       likely to get a defalias on the other.)

The issue at hand was presumably that the buffer's mode (i.e. eww-mode)
is not meant as a "normal entry point".  So it's more normal to have to
keep the name of the handler stable, since that is an explicit entry point.

>       Moreover, instead of (or, well, in addition to) an autoload for
>       the mode function, we now need an autoload for the handler.

But this handler *is* a normal entry point.


        Stefan