On Mon, Jun 16, 2008 at 01:09:55AM -0700, Andrew Lentvorski wrote:

No, it's more about "what does a branch indicate?".

For me, a branch is something that is *not* going to get reintegrated back into the main line later on in the future. And, Lan's examples seem to be pointing in that same direction.

That's definitely something that exists, but it doesn't mean that having
lightweight, shortlived branches isn't extremely useful.

Perhaps they need different names.  Git users tend to call the lightweight
branches "topic branches".

I've experienced numerous instances where branches were useful:

  - lightweight, topic branches.  Basically, any line of development that
    takes more than a single commit.

  - New platform support (mentioned before).  This usually does get
    integrated back into the mainline (or becomes the mainline) once
    everything gets back to working.

  - Upstream integration.  Think working off of say Linux 2.6.24 and then
    spending work to move to 2.6.25.

  - Branch off of an old version that was released to a customer to
    back-port some fixes.  This definitely won't be reintegrated.

Restricting branching to only certain kinds of scenarios seems to limit a
very powerful tool.  Again, though, I never really saw the use of branches
for so many things until using a revision control system where branches
were free, or nearly free (git was not the first).

David


--
[email protected]
http://www.kernel-panic.org/cgi-bin/mailman/listinfo/kplug-list

Reply via email to