On 28 April 2011 00:24, Karl Berry <[email protected]> wrote: > as an example of the GNU system, and as a template for how > GNU packages are written. Does that sound like a fair summary?
As an example of the GNU system, one wants to use only parts of the GNU system to build it. As a template for how GNU packages are built, one wants to use the best, most widely-used tools. > In the case at hand, the added > complications of dVC are not worth the benefits for all projects. The more I use dVCSs, the more I'm convinced they're better in all ways, because they add basic tools useful for both version control (which is as useful to the individual as to the group) and for collaboration that the file system and coreutils simply don't provide conveniently. The increased power for the individual is the best thing: you can't as an individual really use a centralized VCS, at least, only in the smallest, lowest-contention projects, and you can't use them offline at all. The anarchic feel of instant forking encouraged and enabled by github in particular epitomizes the way in which the user is liberated and empowered by dVCSs; several years after first using them and a few after finally being converted, I'm still just starting to get my head round that. I am however convinced that a dVCS (whichever it is) is much more appropriate for GNU: centralized VCSs give power to the administrator, and as such should if anything be discouraged. I think the apparent extra complexity is just that: what's really happened is that rather than having a crude tool for accessing files with history on a single server, we now have a rich tool for sharing file histories. Along the way we have added lots of amazing functionality (git bisect!) that non-dVCSs simply didn't have; that doesn't count as part of the extra complexity of distribution. Much complexity has actually vanished, because complicated (and tedious) manual operations have been automated. But it's only once functionality has been crystallised into commands that it tends to be counted as complexity... None of this helps me choose between bzr & git, though; I'll have a poke about in GNU and elsewhere and a think and see if anything reverses my current leaning towards git. -- http://rrt.sc3d.org
