3 ms·
If I understand this correctly, unlike git-annex and git lfs, this not about extending the git format with special large files, but changing the algorithm for t
by Ericson2314 10y ago
If I understand this correctly, unlike git-annex and git lfs, this not about extending the git format with special large files, but changing the algorithm for the current data format.
A custom filesystem is indeed the correct approach, and one that git itself should have probably supported long ago. In fact, there should really only be one "repo" per machine, name-spaced branches, and multiple mountpoints a la `git worktree`. In other words there should be a system daemon managing a single global object store.
I wonder/hope IPFS can benefit from this implementation on Windows, where FUSE isn't an option.
- manojlds 10y agoThe blog post does mention that some changes have been made to git (in their fork)
- sdesol 10y agoI did a quick comparison of Microsoft's fork and it appears they have done quite a bit with it. Microsoft's fork contains 67,522 commits. The official Git repo contains 45,810. It appears the bulk of the work started in 2010, with significant ramp up of development in 2015. https://gitsense.com/mgit-vs-git/history.png https://gitsense.com/mgit-vs-git/history.png Looks like Microsoft only really introduced about 100 more new files. https://gitsense.com/mgit-vs-git/files.png https://gitsense.com/mgit-vs-git/files.png Microsoft's repo contains 1712 contributors. Git's repo contains 1685 contributors. So it looks 20 - 30 employees worked on Microsoft's fork. https://gitsense.com/mgit-vs-git/mgit-contributors.png https://gitsense.com/mgit-vs-git/mgit-contributors.png https://gitsense.com/mgit-vs-git/git-contributors.png https://gitsense.com/mgit-vs-git/git-contributors.png