Subversion: The new-generation CVS (DevChannel)

Posted May 31, 2004 1:23 UTC (Mon) by gdt (subscriber, #6284) [Link]

Subversion never claimed to be more than a better CVS. That was the project goal. They've succeeded in meeting that, although taking much longer than anyone expected. I've attempted to configure a secure CVS server for updates with anonymous access for reads and been stunned that CVS can't do this without a huge amount of effort (or simply giving up and having two rsynced repositories). So I'm glad to see an alternative to CVS.

You also write of open source development as if it were a monolithic design methodology. This is hardly so; projects vary widely in their methodologies. For example, Samba's "core team" approach might well only see the costs of distributed version control without seeing the benefits.

I've tried arch. A nice idea, but too many limitations for deployment at this time.

I do get frustrated with the "editor wars" aspects of configuration control. Especially since the most useful parts of configuration control are the analysis add-ins, something no open source system seems to have much of. To pick a simple task, rebuild the project each night, e-mailing owners and those that modified failing code with compile and regression test errors. Or, tell me which modules don't have a regression test. Or, who checks in the greatest proportion of failing code. Or, which module has had the most field issues raised against it.

Posted May 31, 2004 6:14 UTC (Mon) by piman (subscriber, #8957) [Link]

Though open source development involves many contributors, it is rarely done distributedly. Almost always there is an official tree, and no others. In rare cases with lots of features in testing, which might conflict with each other, development has actually been done in a distributed manner; this is how the kernel works. But this behavior is rare.

Now, you might say that this is because CVS is so prevalent, and you might be right. But projects were managed like this before CVS, too, and projects that don't use CVS (there still are some!) also usually work this way.

You might also say that although most projects don't use this model, they should. You might be right there, too; but only time will tell on that count. I personally don't think it will have many advantages for most projects.

Posted May 31, 2004 9:45 UTC (Mon) by josander (guest, #19785) [Link]

> Agreed. BK, Arch and Monotone all seem to be revision control
> systems designed to work the way open source development is
> done - distributed

Distributed? Glad you said it (thanks)!

You will find some Subversion based options from the link below (and
don't forget to read the "Dispelling Subversion FUD" doc while you're
there):

http://subversion.tigris.org/project_links.html

Jostein Chr. Andersen

Posted Jun 1, 2004 18:07 UTC (Tue) by dvrabel (subscriber, #9500) [Link]

Which revision control systems a project uses has little to do with how open its development is.