On Sat, Jul 19, 2008 at 11:29:58PM -0700, Tracy R Reed wrote:
David Brown wrote:
Is it really, seriously is that stupidly designed?
How many VCS have you written?
More than 0, less than 1. But how is that relevant? I've been in active
development in several distributed revision control systems. I have a
fairly solid understanding of how most of them are implemented.
The design of most VCS become overwhelmed by the choice of the underlying
data model. How things are stored drastically affects what is easy or hard
to implement. Provided sufficient metadata is stored for changes, most
operations can be done, however.
The last being my biggest gripe about subversion. The underlying model
just doesn't store the right information about merges, and even their
recent fixes don't really make things much better. It doesn't help that
branches and directories use the same namespace. Convention isn't all that
helpful to a program trying to determine if part of a path is a branch or a
directory name.
No wonder people comment out blocks of code if you can't do something as
> simple as reverting an old change.
You can revert an old change. Quite simply.
How simple is it? Is there a single command to back out changes, or does
the user have to track down which files are modified by the change and do
it to each. If there is a single command, it makes it easier than P4.
There isn't any particular reason that SVN couldn't have an easy revert
command. It's just that based on the commands that are available in SVN,
the developers seem to have a very different idea than I do as to what
revision control even _is_.
David
--
KPLUG-List@kernel-panic.org
http://www.kernel-panic.org/cgi-bin/mailman/listinfo/kplug-list
