Erik de Castro Lopo wrote: > > If anyone is interested in how I'm doing this, I can make my Bzr repo > public. I should also mention that none of this is Bzr specific > and it would be trivial to move this to any other modern VCS.
Just as a related data point on how small git repos are... You don't need to store tarballs in a git repo; rather, you'd import the content of them as upstream makes a new release. The only tarball-related git thing you might do is use pristine-tar, which writes a small bit of metadata that allows you to generate a bit-identical copy of the original tarball using only what's in the repo; but that's an extra feature past what your setup had anyway, and those pristine-tar bits are usually less than 10K anyhow. My Debian convertible repo's .git is 136K. If I look at my own HDBC tree -- the one with every single commit since HDBC began -- it's 408K. That's 314 commits. The HDBC 2.1.1 tarball is 31K compressed, 143K uncompressed. Each revision of conertible added just over 1K to the pristine-tar branch. So we're talking really small numbers here. This is why I say git is really efficient on disk. -- John -- To UNSUBSCRIBE, email to [email protected] with a subject of "unsubscribe". Trouble? Contact [email protected]
