Re: Re: Subversion for backups?

[Adam Spiers <orgmode@adamspiers.org>]

On Thu, Aug 27, 2009 at 05:27:40PM +0000, Robert Goldman wrote:
> FWIW, my feelings are the exact opposite of Matt's --- I find the added
> complexity of having multiple repositories to manage and the general added
> complexity of git quite unwelcome.
> 
> If you are trying to keep multiple machines in sync, I find that a centralized
> repository scheme is far simpler:  all you need to do is manage the relationship
> between your working copies and the repository, and that relationship is a very
> simple one.

This centralized model is also entirely possible with git.

> If you have a distributed revision control system and multiple different
> repositories, you must manage the relationship between the different
> repositories and the relationship between those repositories and your working
> copies.  Because of the peer-to-peer aspect, this is a complex quadratic mesh of
> relationships to manage, instead of the simple linear relationship you have with
> a centralized repository.

It's not quadratic - it's still linear because it's quite unnecessary
to have every repository interacting with every other.

> I would say that if (1) you generally are connected to the internet, with only
> minor intervals offline and (2) have access to a hosted svn repository (so you
> don't manage it yourself, and so that you can use the simple https protocol
> instead of fussing with ssh tunneling), then you are likely to find svn much
> simpler.  I am fortunate that both of these hold for me.

I disagree - I think using git with a centralized model provides the
best of both worlds: simplicity but also all the nice benefits of
decentralization such as offline commit and history access,
intelligent merging etc.  Lots of people do it this way, e.g.

http://feeding.cloud.geek.nz/2008/12/setting-up-centralied-git-repository.html