For anyone who missed it: I gave a lightning talk during the morning session here at PyCon today, following up on my earlier questions about distributed version control. Slides are available as a PDF.
Next up: attempting to live-blog the ORM panel.
Comments for this entry are closed. If you'd like to share your thoughts on this entry with me, please contact me directly.
I liked this presentation. But I’d add one caveat: You’re talking about mature projects that started with CVS or Subversion and later switched to DVCS. Of course they still have processes and social norms that lead to centralized workflows.
When I compare that to the activity around brand new projects that are born every day on, say, GitHub, then it seems that there really are different ways of doing things. When I spy a new project that interests me these days, I’m far more likely than before to clone the code, experiment with it, and publish anything interesting I come up with - even if it’s not necessarily of interest to the original author. In a truly distributed project, I’m much more likely to scratch my own itch, because I know that my patches will never just die a quiet WONTFIX death. At worst they’ll live on in my personal fork for anyone else with the same hopes and dreams who happens to stumble by.
And then there’s Linux, which is a world unto itself. It’s far more complex than just “what Linus says it is.” It would take a team of anthropologists to properly catalog all the ways kernel patches flow from their authors through the many trees with their different roles.
To be honest, successful projects always are centralized. They, at the very least, have a gatekeeper, which is still better than having a bunch of commiters.
The difference isn’t in the technology, but in how people can and will use it. At the very end of the day, Django is no different to Rails, from a 10k feet view. It makes website backends, right?
Well SVN and Git, both keep code history (or file/contents history, to be more precise). People use it for the same thing, in different ways.
This presentation appears to focus on the fact that the official point of distribution is centralized. I always understood the point of DVCS to be the capability of peer users to exchange changesets without requiring a shared central repository.
The interleaving of locally-committed changesets (with some shared history) into an aggregate branch is functionality that your proposed “super-svn” doesn’t seem capable of performing.