On 6/14/07, Igor2 <[EMAIL PROTECTED]> wrote:
> I am not against using git, but I need to protect SVN here. One needs to
> see both the adventages and the drawbacks.
>
> With git, I will need to worry about disk space, because checking out a
> repository will download and store much more than an SVN (or
> CVS) client. If I'm an active developer, it may be a good idea, having my
> own history, but if I'm only an user interested in the latest revision,
> this may be a pain. Especially if I don't have 160 gig disks. Of course,
> with newer git versions there may be a way to get the latest release only
> using extra command line arguments, but an "svn co
> svn://something.org/project/trunk" may be simpler for users than a long
> command line. I am not sure (yet) how git works, does the same happen
> with updates? Can I use an "svn up" type command and only have the latest
> revision and no local copy of the whole history?

I don't think git can do this... because git is implemented in a
distributed point of view and this was made to make it easier for the
developers to keep their work while without internet connection or
whatever.

>
> The other important point is the distributed vs. server-client
> approach. Everyone having his own repository is really nice because more
> develoeprs can start experimenting with the code and they can share their
> modifications easier. However, the "boring" server-client setup with
> it's centralized approach can send an email or irc message on any commit
> to a mailing list so developers see what other developers are working on
> in a branch. The repository is usually hosted on a high bandwith server so
> everyone can get the files fast. The serve may have a stable network
> connection and high uptime. Comparing this to having to connect a random
> developers home computer behind an adsl fo a specific branch has
> drawbacks. Or if the developer pushes (?) his branch to a central server,
> we are back at client-server approach. Or does git have a trick for this
> as well?

git can do that also... you just need to give execution rights to the
sample post-commit hook script. If you checkout the mailing list
MM-commits (related to the linux-2.6-mm tree maintained by Andrew
Morton) you'll see every single commit is posted on that mailing list.

git works like this... you clone a repository... work on that
repository... commit your changes locally and then you
git-format-patch your changes and after that you git-send-email all
the patches to the mailing list... on the central server side, the
maintainer just git-am the mbox file with all the patches in it...

Of course you can push... but the only one who can push directly to
the central server is the maintainer himself...

>
> People usually say that svn is centralized and you can not work offline
> for long because you can't save little commits. Of course, with vanilla
> svn it's true. However, there are ways to change this, for example svk, so
> one can have a distributed svn.
>
> Finally, I think it's a good idea to keep read-only CVS mirror if the
> project moves to git (or anything else).

Even if it moves to git you can export git commits to svn or cvs
commits using a simple command, this way would be easy to keep a
mirror in cvs.


-- 
Best Regards,

Felipe Balbi
[EMAIL PROTECTED]


_______________________________________________
geda-dev mailing list
[email protected]
http://www.seul.org/cgi-bin/mailman/listinfo/geda-dev

Reply via email to