[
Date Prev][
Date Next][
Thread Prev][
Thread Next][
Date Index][
Thread Index]
[
List Home]
|
Re: [udig-devel] subversion
|
Bart van den Eijnden wrote:
Hi list,
uDig is the only project that I know of that uses Subversion instead
of CVS.
Let me introduce you to another one - geotools
http://geotools.org/index.php. UDig is using subversion (svn) to allow
us to cooperate more closely with the geotools project.
Geotools chose subversion primarly for its ability to maintain version
information cross renames. You must also take into account that the
alternative was CVS on SourceForge (painful).
As long as you are shopping you may want to look into a tool called Arch
(it has a formal mathmatical model associated with its versioning system
- kind of like relational algebra/tuple arithmitic and its association
with modern databases).
My own limited experience with svn is that the Eclipse plug-ins are not
ready for prime time, it has been a joy to work with on the command line
though. Branching is at least easier then cvs.
But one of the best things to do is go look into the archive of
pragprog@xxxxxxxxxxxxxxx they recently had a nice thread on all these
issues.
I will cut&paste a bit for you:
Dave Thomas (dave@xxxxxxxxxxxxxxxxxxxxxxx):
I've been playing with Subversion recently. It has very much the feel
of CVS (intentionally), but fixes some of CVS's shortcomings. For
example, commits are atomic, it versions permissions, and it handles
renaming, moving, and copying files. It also has an interesting
alternative to tagging and branching.
The down side is that it can be a bit of a pig to install a server: it
needs Apache 2, a particular version of Berkeley DB, and a number of
other libraries. However, once built, it seems to work well, and I'm
entrusting some of my play projects to it.
Tony Bowden (tony-pragprog@xxxxxxxxx):
A very interesting choice is Aegis[1], which is a free (speech and beer)
CMS that also provides for:
enforced testing
a change can't be integrated unless it
passes its tests,
passes the regression test
fails its new tests with the current baseline
[i.e. when fixing a bug you make sure it can't ever come back].
peer review
someone else has to approve the code before putting it live
automated build
can be used for almost anything
we use it for checking local 'policy' decisions by analyzing the
code, and reformatting the code through a pretty printer to avoid
holy wars over style, and ensure a consistent code style across
the entire project.
Tobias C. Rittweiler (tcr@xxxxxxxxxxx):
GNU Arch. It's [going to become] the best distributed revision
control system. Especially its branching facilities are par exellence.
If you ask me, Arch's gonna kick ass bitkeeper, and it's already way
better (and not only because it's FLOSS).
However, arch has quite a steep learning curve, it usually takes 3,4
week before you think arch. Also, even though the windows support has
been greatly improved over the last week (it now does run on it with the
unx tools [or how they're called])
ANd back to David Thomas:
Know know: I'd really like to like arch, but it really needs to lighten
up a bit first. Right now it mixes too much of the author's philosophy
in with the functionality for my taste. I'd rather have a
lighter-weight tool which could be used in differing ways. It also
needs to have someone go through and totally replace the CLI: it is
ridiculously verbose as it stands, and it seems like its far too easy
to make mistakes.
Having said that, these are superficial issues: I agree that underneath
it all arch has promise.
Subversion _had_ promise, but I'm not sure it's going to deliver on it.
It's still far, far to complicated to install (this drive to use the
latest and greatest of just about every imaginable library is more
widespread than just subversion, but in the case of subversion I feel
that it really limits the takeup).
I'm not sure there is _any_ clear winner in the version control world
right now. Older system have problems, and newer ones seem adolescent.
(But I've love to be proven wrong)
