Many open source projects uses git & github those times. Have you considered moving from svn to it? You can import svn history:
Git? As you can see, our development model is definitely centralized. A few more reasons to use SVN:
- ease of forking – actually a drawback for us, we already suffer from Android port (nobody notices it’s a fork and they ask us for support)
- long, unreadable, incomparable revision numbers
- bad support for Windows, at least last time I checked it TortoiseGIT was considered significantly worse than TortoiseSVN (I don’t understand why anyone would like to use cmd line for version control system)
- ease of creating local branches (actually an advantage but the only one I can see).
How about mercurial then?
- nobody makes you support anything, forks usually maintain themselves (I don’t consider it an argument for whether or not to use dvcs)
- you can use your ‘regular’ svn-sorta-style revspecs (see: blogs.gnome.org/newren/2008/03/3 … pecifiers/)
- quite good windows support, see visualhg.codeplex.com/
- branches, but this you already know
the list above is just to show you that each your argument can be trumped, but that’s not all:
- see them google folks recommend hg (or dvcs’ in general): code.google.com/p/support/wiki/DVCSAnalysis
- the speeeeed. ever tried to fetch a log or blame for a file?
- clear revision graphs
- can be used offline
- even m$ themselves think it is good blogs.msdn.com/b/codeplex/archiv … oject.aspx
now don’t get me wrong, I’m an avid mercurial zealot, but I just see no reason to use that PITA svn anymore, especially when proper tools are available.
sf supports hg, though it uses quite dated version.
Where are those svn-sorta-style revspecs? I can’t see them – only sha1 hashes appear.
tortoisehg uses it by default http://dl.dropbox.com/u/1323345/mercurial.png
don’t know about web interface though. surely, you can use svn-style revspecs manually like this: bitbucket.org/janisozaur/vcmi/changeset/2516
will see into that later on
you see, the problem with svn-style-revspecs is that they’re specific only to particular clone. with move to dvcs came offline commits and there is no centralised authority that issues them, as that would bring us back to the start.
I know they’re somewhat useful when comparing revs, but when I switched from svn to something more handy (be it git, hg, bzr - used them all) I’ve realised that they are not that much needed. You just usually copy them around and don’t have to memorise each one and since you have a repo lying around - you can always check. point’n’click style for you windows users is always available.
After switching I haven’t once regretted and I miss nothing, this includes svn revspecs. Trust me, you don’t need them, it’s only illusion.
There is a workaround - you can setup a service hook that would apply bookmark ‘svn revno:xxx’, but like I said, this is only a workaround and not a solution to an imaginary problem.
The real advantage (one of many) over svn are pull requests - you can have community do some coding and they don’t need to have svn commit rights. when you get a pull rq, you can see the commits, review code (many hostings allow to comment in-line) and apply it to your ‘trunk’ with just one click.
I urge you to try hg or git and then conclude. Just please don’t start with negative attitude but be neutral. This might be a new thing for you and might have some learning curve, but is is very flat one.
couple useful links from google:
git crash course for svn users: git.or.cz/course/svn.html
mercurial for svn users: wiki.pylonshq.com/display/pylons … sion+Users
you can also read many of 'why i switched from subversion to hg/git’
one quite popular run-down of git features compared to others is whygitisbetterthanx.com/
SVN feels fine for me, as long as it works (it certainly does). However, recently Sourceforge introduced totally screwed stats system, which is not useful at all and generally annoying.
But bitbucket link you provided is a tragic"
Totally illegible and useless mess.
I’m surprised you mention the stats system. you mean the graphs that show repo/files traffic?
I see no point in bothering, I’d even assume that you don’t see it quite often as it requires a trip to sf project page and since you have this forum and separate wiki - what for?
it was done completely automatically, there is no user mapping. other things are pretty generic: link to parent, which branch, commit msg, the actual diff… how is that illegible and useless?
Trac certainly is more readable and provides simple, yet meaningful info. No hash (?) codes filling half of the screen.
Ohloh is something than I found and which looks better than what we have now, but no one proposed it yet.
Warmonger, ohloh (I think) is not a repo/hosting per se. They mention it in the header:
It is merely a stat tool, but if you want stats, to each his own.
You should just try to use git. Advice to use dvcs is only a care about you and to help other people to make contribution to this project.
As mentioned above your args are not convincing:
You can continue to use your model there is no any problem. In company where I work few projects recently moved from svn to git. We continue to use the same model and the everything is good.
I think that such maintenance should be continued by author of a fork. Actually it’s not so hard for fork author as well as for you or somebody else. The solution is just to update the fork with your main development branch and make necessary changes. Also in my company I support our C++ ORM library for windows, win ce and linux. Actually problems are mainly related to code organizing but git gives a big help to very easily keep current, development and merging (features and platform related changes from other projects) branches.
Please, forget about tortwisegit. It’s awful, ugly, buggy and moves you from git ideology. There is an excellent tool for windows - GitExtensions. It covers all daily operations and has a nice GUI.
It’s a great feature! What steps somebody have to do to try make changes?
- I would like to create at least local repository, of course I will use git.
- Create test branch and try to make small changes to see effects
- Try to implement some feature
- Well, I’ve become enough familiar with project and would like to send changes to you. What should I do? I’m sure that with git it’s easier and safer.
Versioning is also not a problem. We use script which relies on git describe. There are few drawbacks but actually nobody has problems, we even have not problems for versioning for ‘test’, ‘pre-release’ and ‘release’ in one repository.
PS:
During the moving we faced with few problems. But the main problem is to understand the concepts of git by few team members. The root is lazyness. Of course it takes some time and everybody makes mistakes. So don’t worry if you guess that something wrong, in many cases it means that you should think about it a bit more.
Heah, git\hg would be a win. It’s really hard to contribute to open source project hosted in svn. For example I want to play a bit with code, but I first of all could not make my own commits (unless “git init” 
And even when I would have done something there is no easy way to share my changes. I need to make patches and send them via mail I guess. In github\bitbucket I can make a pull request. And everything this using nice interface where pull request could be accepted\declined, commented what to fix to get accepted, etc. Also github\bitbucket have nice built-in issues trackers and wikies.
So guys you really need to have one more look at git\hg and give it a try.
Right now I’m using Github for a big (24 people) university project. I’ll see how it works actually. Shame it has a nice graphical tool for Windows but not for Linux.
So still no willing to move to github/bitbucket?
Same people, same tastes…
So it’s probably unlikely?
At least now. Maybe some git lover will do a great job for the project and convert us :).