From: Jeff King <peff@peff.net>
Date: Fri, 17 Dec 2010 21:21:00 GMT
Subject: Re: [PATCH] completion: add missing configuration variables
Message-ID: <20101217212100.GB11511@sigill.intra.peff.net>
In-Reply-To: <7vei9gi0y3.fsf@alter.siamese.dyndns.org>

On Fri, Dec 17, 2010 at 12:44:36PM -0800, Junio C Hamano wrote:

> > All variables are case-insensitive. The config parser down-cases them,
> > so all code should treat tham as all-lowercase. However, we tend to
> > document them as camelCase for readability.
> 
> Careful.  The second level of three-level names is supposed to be usable
> for user defined names (e.g. names of branches and remotes) and I think
> the completion code should not downcase it.

Yeah, good point. It is not even "second level" in the parsing code,
though, since that code sees:

  [first "second"]
    third

and not

  first.second.third

IOW, the second one is ambiguous in the face of an internal dot.  So if
I had something like:

  [Foo "Bar.Baz"]
    Bleep = true

then "Foo" and "Bleep" are case-insensitive, but both "Bar" and "Baz"
are not. So in terms of dotted components, the rule is "the first and
last dotted components are case-insensitive, and middle ones are
case-sensitive". At least with respect to my simple test of calling "git
config x.y.z". I'm afraid if I look too deeply I might discover some
inconsistency. :)

-Peff