[Noisebridge-discuss] [dorkbotsf-blabber] Embedded code - version control experience?

Josh Juran jjuran at gmail.com
Sun Feb 19 21:30:17 PST 2012


On Feb 19, 2012, at 2:21 PM, Larry Edelstein wrote:

> I've worked in software for a long time now, and feel qualified to  
> speak on VCS. It's changed, and you valiant garage-dwelling  
> embedded guys and dolls (please allow me my romantic image of  
> y'all) shouldn't miss out on what's new.
>
> The "distributed", peer-to-peer version control systems like Git,  
> Mercurial, and Bazaar are much better than dinosaurs like CVS, SVN,  
> SourceSafe, etc. for just about any project.

CVS is a long-obsolete, inadequate, and overly flawed tool, but it  
doesn't deserve the dishonor of being mentioned in the same sentence  
as SourceSafe.

>  But if you make the investment to get used to distributed VCS, you  
> will never go back.

Agreed.  I went from CVS straight to Git, and now I find SVN  
unbearable.  If I need to work on a project in SVN, I download a  
source tarball and check it into a local Git repo, hack on it, and  
send patches to the mailing list.

> When you use one of these new-wave VCS, you store your code in a  
> repository a server, like you're used to, but you and anyone else  
> who works with that code replicates most or all of that repository  
> on their development machine. (Now that we all have plenty of hard- 
> drive space and low-latency high-bandwidth networks (er, except  
> Carl), replication is cheap.)

Especially if the project has SVN "branches" or "tags" for each  
release.  Try checking out the nginx SVN repo, for example.  You get  
3 *gigs* of data; the actual checkout is only about 5 megs.  And the  
checkout takes forever.  Git keeps your entire history in each client  
repo as well, but does so in a way that's efficient both to transfer  
and to store on disk.  (For one thing, it doesn't keep multiple  
copies of the same blob of data.)

Josh




More information about the Noisebridge-discuss mailing list