Re: [dev] [suckless] Migration to git

From: markus schnalke <meillo_AT_marmaro.de>
Date: Tue, 27 Nov 2012 11:36:13 +0100

[2012-11-27 12:03] chris hall <followingthepath_AT_gmail.com>
> On Tue, Nov 27, 2012 at 8:50 AM, markus schnalke <meillo_AT_marmaro.de> wrote:
> > [2012-11-26 16:47] "Roberto E. Vargas Caballero" <k0ga_AT_shike2.com>
> > > - History rewriting: git rebase or git filter-branch
> > (btw: That's the worst feature a version control system can offer.)
>
> I disagree, IMO this is one of the most useful features of a vcs.
> This isn't usually rewriting history on a global scale (that is
> possible, but usually very dangerous) but instead only altering my
> local and unpushed commits.

(Sounds a bit like hard links to directories ...)

> I use this for 2 different workflows:
> In one I have a WIP commit that I add to every time I am closer, and
> then break this up before pushing.
> The other I instead make a commit every time I get a step closer, and
> then squash them into logical commits before pushing. This gives me
> version control within an individual feature, although I guess you
> could achieve the same thing with branches and manually cping files
> around.
>
> I find a vcs constraining if it doesn't allow me the use of these
> workflows, YMMV

History rewriting is no necessity for these workflows. Cloning of the
repo (which replaces branching) allows you to do the same. You are
always able to clone any arbitrary revision. You guys just fear cloning
a repo ... (It is only a performance issue if your project comprises
hundreds of thousand lines of code and a huge number of changesets,
which we surely don't have.)

It's the ingenuity of Unix to have just one kind of fork(). Why can't
we map this concept to DVCS and have just one kind of cloning?


meillo
Received on Tue Nov 27 2012 - 11:36:13 CET