From: Jeff King <peff@peff.net>
Date: Sat, 07 Feb 2015 04:52:20 GMT
Subject: Re: BUG: 'error: invalid key: pager.show_ref' on 'git show_ref'
Message-ID: <20150207045219.GA15548@peff.net>
In-Reply-To: <xmqqk2zud6b0.fsf@gitster.dls.corp.google.com>

On Fri, Feb 06, 2015 at 02:27:31PM -0800, Junio C Hamano wrote:

> Jeff King <peff@peff.net> writes:
> 
> > A list of enum-like values where we are OK confining the names to the
> > alnums is OK to use as an unbounded set of key values. Just like we have
> > color.branch.*, we just pick a name within that syntax for any new
> > values we add (and that is not even a burden; alnum names are what we
> > would have picked anyway).
> 
> I would say that color.branch.<slot> names are very different from
> subcommand names.  The latter is exposed to the end users who do not
> have to know that they can be used and must be usable as config
> keys.

Yeah, again, sorry if I wasn't clear. That was the same contrast I was
making. Of the examples given in this thread, color.branch.<slot> and
fsck.* names are in one boat ("OK to give them configuration-friendly
names, they are just a list") and arbitrary commands are in another.

-Peff