How to use Mercurial?

[Jon Ribbens]

On Thu, Apr 08, 2010 at 11:10:24AM -0400, Warren Postma wrote:
>    It was precisely my mindset, until my own experience beat it out of me.  I
>    find that my experience was shared by many many others. That Joel guy
>    found the same thing -- see [1]http://hginit.com/ and read the parts for
>    refugees from centralized version control systems (svn and cvs).

Yes, that site is the reason I finally decided to look for something
more modern than 'cvs' in the first place.

>    It sounds to me like you have the following situation:
> 
>    1. More than one project (set of source code or input files) per
>    repository. This is a rather un-mercurial style.

Not really. We only have one "thing" per repository. What I meant by
"project" is that one person may be working on "Thing A" to add
feature X, and another person may be working on "Thing A" to add
feature Y (and indeed the same person may also be working on a
longer-term project to add feature Z).

>    2. Only one local working copy (also rather un-mercurial, I think).

No, we have one local copy per project (using the meaning described
above).

>    You prefer everybody work from the latest source code, perhaps to reduce
>    merging.

As far as I can tell, keeping up-to-date *increases* merging. But if
we have fixed bugs or added features in the main live repository then
we want the local repositories to have those changes applied so that
the projects are working against the "real" version of the code not
some old out-of-date version.

>    What if your individual developers have two local working copies,
>    at a minimum:
> 
>    a.  My Pile Of Doggie Doo.
> 
>    b.  My Copy of the Central Pile of Doggie Doo.
> 
>    And then I move from pile A to pile B when I feel like merging, but not
>    when I don't.

Sorry I don't really understand what you mean. I thought about having
it so that each developer has their own "central" repository which
pushes/pulls from the "central central" one, and their "working"
repositories push/pull from each "developer central" one, but it
seems to me to add a lot of complexity for absolutely no benefit
(well, except for the utterly trivial benefit of repositories using
hard-links to save disk space).

>    Am I understanding you?

Not entirely I'm afraid ;-)