Re: distributed version control systems
On Thu, May 18, 2006 at 05:42:16PM +0200, Geert Stappers wrote:
> On Wed, May 17, 2006 at 10:20:12PM +0200, Sven Luther wrote:
> > Imagine i upload a new nobootloader version, with some changes in it. I either
> > upload it directly, without revision system, or with my own shadow copy of the
> > d-i svn like above.
> > Now, someone else needs to modify nobootloader. He is not aware of my changes,
> > commits to the d-i svn, and uploads the package. My changes are lost.
> > Next time i upload my changes, i may well not notice that there was another
> > upload, and the other changes are lost. (Well, probably not, because i will
> > have some svk based tool to merge those changes into my tree).
> > So, we end up in a mess, because there is no more only a single authoritative
> > (or even juste a single) copy of the repository for the package.
> > This is the reason why i am arguing against the current proposal which doesn't
> > restore my svn d-i access, and why i have not even tried to do any d-i work
> > since then.
> And there is the possiblity to send `svn diff` output to this list.
Sure, but it is inconvenient, and there are other things to do.
But even then, there could be a lag from the list to it being checked in, or
whatever, the potential for making a mess is always there, as much as you try
If we chose a centralized repository like subversion, then the only logical
conclusion is to keep all changes in that repository.