Re: [Gnu-arch-users] subversion vs. cvs (vs. arch) over on pgsql-hackers

[Pierce T . Wetter III]

On Mar 25, 2004, at 5:28 PM, James Blackwell wrote:

> >         * Decentralized revision control makes things harder.
> >
> >
> >         There was also my argument that distributed systems support
> > centralized development, but not vice-versa to which someone responded
> > ``that is what svk does.''  I'm not at all familiar with svk, but it
> > describes itself thusly:
>
> Though you don't hear about it much, centralized support is built right
> into arch. Just have everybody commit to the same archive. Everything
> will just work.
>
> I use this when working with people on web-based projects. Because of
> the nature of the beast, web based projects demand centralized
> development. Arch works absolutely *great*.

 Yeah, I kind of think arch would be easier to grok if the 
documentation was broken into these pieces:

  Chapter 1: Centralized Development with Arch

     Basically, how to use arch instead of cvs

  Chapter 2: Branching with Arch

     How to branch well.

  Chapter 3: Decentralizing your development

     How to use arch with some typical development use cases that show 
the
     power of "unblessed" repositories, like laptop development, 
distributed
     development, local branches of large projects.

  I think people who are attracted to arch are more interested in 
chapter 2 and 3 then chapter 1, but its still better to split up the 
docs that way, because chapter 1 will be much shorter and simpler then. 
For instance, with a central repository, you don't really have to 
understand what archives are because all that 
address@hidden stuff is just "stuff you have to type 
once" in the same way that you might type:

setenv CVS_RSH `which ssh`
cvs -d :ext:address@hidden:/Repository checkout Projects

 once, but after that its all "cvs update" and "cvs commit".

 Chapter 2 could then go more into the nuts and bolts as far as 
specifying branches and such. So that's CVS plus good branching, and 
how to use the star-merge command.

 Then Chapter 3

 This can explain more complicated stuff like set-tree-version, 
mirroring archives.

 Pierce