Cool news: John Wiegley made a new release a few hours ago, my patch is no
longer relevant, will make another one, which just bumps the version to
2.4.1.

> Normally, we try to install optional packages per default.
> Sometimes, it's not so easy and one has to weigh up...
>
> Maybe the use-case for use-package is so that diminish is very often
> needed?  Or for too many users, it would be "useless" (or break
> something), w/o it?
>
>
> Imho (I don't have commit superpower), you need a reason to remove an
> input (and here, this means provided/out of the box experience).  Or at
> least make a comment in the source _why_ it isn't needed etc...

I will give you a simple example: Someone uses delight (a diminish
alternative, which also support major modes) instead of diminish and
they would install emacs-use-package and emacs-delight, but he will also
get emacs-diminsh as a dependency of emacs-use-package. To remove it
they will need to inherit emacs-use-package and update
propagated-inputs.

I think installing emacs-use-package and manually adding emacs-diminish
is easier and more fair than installing emacs-use-package+emacs-delight
and removing emacs-diminish from dependencies of emacs-use-package.

I totally agree, that diminish is a great package with a great story
inside its sources, but don't think it should be installed by default
with use-package.

I'll make a separate patch for removing propagated-inputs and related
discussion.

Bump patch in the attachment.

-- 
Best regards,
Andrew Tropin