Emacs packaging guidelines (was: Re: Fedora minimal installations)

[Michael Schwendt]

On Wed, 29 Jul 2015 20:12:56 +0100, Jonathan Underwood wrote:

> >> >> https://fedoraproject.org/wiki/Packaging:Emacs for more detail."
> >> > And once again the wording is weak. It says "should". Three times even.
> >> What's wrong with the word "should"? What am I missing? Seems like the
> >> usual use of this word to me?
> >
> > Presuambly this: https://www.ietf.org/rfc/rfc2119.txt

Sort of.

> OK, so that's exactly what I intended when I wrote the text of the bug
> report (I had that definition in mind) - I thought there might be
> unforseen valid reasons not to merge the emacs- subpackage into the
> main package, and hence didn't use "must". So, I am still not seeing
> my error here but I would like to, so I can avoid it in future...

I am not a native speaker of English, but indeed there are stronger
forms, such as "ought to", "must" and "have to". The latter two are
more clear about whether the packager is permitted to create an emacs-
subpackage as a matter of freedom. On the contrary, "should" implies
freedom, i.e. the packager "need not" ship Emacs Lisp files in the
base package and "need not" add a dependency on emacs-filesystem,
but may create an emacs- subpackage instead.

It's the same with the "multiple directory ownership" guidelines, which
don't mandate that packager _must_ require -filesystem packages, if
available. They say "should" and people read it as a recommendation
instead of a strict rule/policy.

When dealing with [new] packagers in the review queue, ambiguous wording
can be a pain.