Re: Revision control

[olafBuddenhagen]

Hi,

On Thu, Jun 19, 2008 at 03:11:02PM +0200, Arne Babenhauserheide wrote:
> Am Donnerstag 19 Juni 2008 00:46:55 schrieb olafBuddenhagen@gmx.net:

> > > Once the Hurd has reached 1/100 of the RCS traffic of Linux, a
> > > switch to a powerful tool might be feasable.
> >
> > That's like saying that it's only useful to run GNU/Linux on very
> > large servers, because others don't need the power...
> 
> No. 
> 
> It's like deciding to go on a trip through the country together and
> saying: "We don't need racing cars which run only on a small subset of
> the roads and have fancy controls most of us have to learn from
> scratch. We want to drive together to a place we don't necessarily
> know already." 
[...]
> The difference you name would in my opinion be better described by
> saying, we only need OpenBSD on the most critical servers, because for
> the others, its gain in security isn't worth the effort to set it up
> correctly and the time to learn its quirks. Would you use OpenBSD in a
> university as shared working server for people? 

Both your examples miss the point: These are not suitable for everyday
usage, because they are extremely specialised. Git is not. On the
contrary, git is extremely generic -- which is precisely what I like
about it :-)

> One reason for wanting to be able to drive many different roads: The
> informatics professor from whom I learned about the Hurd uses Windows
> on his presentation machine (on which he also works quite often), and
> he wouldn't be able to use Git efficiently in that environment. 

"Frankly, my dear, I don't give a damn" ;-)

It's nice to hear that there are computer science professors talking
about the Hurd; but I don't buy the argument that being able to easily
access the repository from Windows matters for Hurd development in any
way.

> > There is probably a difference between a project with a single
> > contributor, and a project with more than one. But otherwise, the
> > size or activity of a project has very little influence on what kind
> > of things the developers need or want to do regarding version
> > control.
> 
> Just look at the workflows in Linux and in other projects. As far as I
> know, there is a huge difference. 

Every project has a different workflow -- both small and large. Git is
nice because it serves *all* kinds of workflows; it is extremely
flexible in this way. Again, that's what I like about it...

> But the Hurd isn't in the position to force devs to learn new ways to
> contribute (not that I like forcing people to do something they don't
> need). It needs developers to be able to jump in easily, and every
> programmer lost on that way due to an infight with Git would hurt
> quite much. 

Sorry, I just fail to believe that any serious contributor will be
detained by "infight with Git".

> And Xen, xine, SymPy, OpenSolaris, OpenJDK, NetBeans and many others
> aren't really one-dev projects, so they should be enough hard proof,
> that Mercurial works well for bigger projects. 

And many equally large projects still use CVS, or SVN. So what exactly
does that prove?...

> But in the end, much about the current distributed version control
> systems depends on gut feeling, personal preferences and the needs of
> the specific project, since they are quite similar, and there is one
> big shiny light at the end of the decision tunnel: 
> 
> Whatever we choose, the technical side of switching the files and all
> history from one to another is very easy, since there are simple
> automated tools for the task (Git to Mercurial as well as Mercurial to
> Git). 

Indeed :-)

-antrik-