Why not just use BitBucket? It's free and reliable and is meant as a repository manager. And trusted like GitHub, but BitBucket also gives you unlimited public AND private repos. Did I mention free?
Considering superior alternatives like mentioned BitBucket, I don't think I will use this project any time soon. But I'm very glad that some has created it. For me it demonstrates that git is flexible enough to support hosting/sharing solutions that are not standard today. I take it as a sign of git being a least a bit futureproof.
But aside from that, do all hacks have to be practical?
I don't know about BitBucket, but GitHub has suffered downtime before. If you have a second remote on Dropbox, you can also get to your files there and wherever it's automatically syncing them should your VCS go down.
No, that's what you get with the GitHub student account (free for any student with a .edu/school email). BitBucket offers unlimited repositories for free.
Those services add:
- Access to GitLab Enterprise Edition (EE)
- Next-business day support
For a little more:
- 24/7 Emergency Support
- Live installation and configuration assistance
- Live upgrade assistance
- High Availability
- GitLab CI Support
An for a lot more:
- Prioritize features important to you
- Best practices training
- Dedicated Account Manager
By not doing it forever. They are competing with two entrenched players, so this is probably a move to gather a larger user base. They are probably going to limit private repos for new users at some point.
We think this market is going to end up like email, there was hotmail with 2MB but no private user is paying for unlimited email storage now. Our offer is like Gmail, except long term we want to make money from an application marketplace instead of advertising.
Once they the education.github.com thing with a proper campaign and starter pack, they received way too many applications than they could handle.
It took them a long while to sort through the backlog, and its always increasing. My education plan was renewed just a few months before I graduated and I'd applied for it almost 8 months back, I think.
Looks good, I had previously used Dropbox for personal repositories, but now I might try using them for shared ones.
It seems from a quick read of the design document [1] that the important thing here is the updating of refs using a compare-and-swap, ensuring that the synchronisation issues that typically occur with using Dropbox as a remote cannot occur. Objects are always* safe from this since the file names are SHA1 hashes.
Edit: The only downside I can see with this is that the remote will never have garbage collection take place, so unreachable objects will continue to take up space indefinitely.
Yes, you're right, there is no garbage collection on the remote. (I should probably add that to the design doc.)
I don't know a way of doing it client side that is concurrency-safe. Perhaps I'll write a separate script that will do remote GC, but will require that you manually ensure that nobody is modifying the repository concurrently.
Also since packs are not used (the design document says it stores all objects as loose objects) there is no delta compression. This won't be usable for large repos with lots of history.
Yep, no packing and no delta compression. I haven't yet thought of a good way of doing that entirely client side. (If you can think of anything, please let me know!)
The target use case is small personal and group projects.
With the current design, it won't scale that well to large repos with lots of history. The initial clone will take a while. However, individual fetches or pushes after that shouldn't take that long unless there is a ton that changes.
Just curious... Given that we have gitlab, bitbucket, assembla, etc. which provide free private git hosting and even more services if you don't need private repo - why would anyone host on dropbox?
- Mobile apps: ability to browse/create super easily from your iPhone or whatever using any app that integrates with Dropbox
- Storage space: though not always written in stone, most git hosts expect code and prefer you keep your media, if its heavy, elsewhere. Users of git-annex, largefile, etc. benefit from this.
- Collaboration: lots of people use Dropbox, not all of them need to use git (or at least, not all the time). See above.
Unfortunately, after reading the readme I see this: "Do not directly interact with Git repositories in your Dropbox folder - always use git-remote-dropbox. If you're using the Dropbox client to sync files, it's a good idea to use selective sync and disable syncing of the folder containing the repository to avoid any unexpected conflicts, just in case." So it's not as rosy as my above points hoped for.
Seeing all those "why not use X instead" makes me wonder why it appears to be a binary option? Git lets you use ALL of them. You can push to multiple remotes and even automate things with hooks if you need to. Perfect for backup and redundancy in case any single remote is unavailable...
Or you know pay $5 and host your git repositories at Digital Ocean or some other VPS provider. I honestly don't get this obsession of tacking on git on top of things like Dropbox. If you're technical enough to understand git then you're technical enough to set up your own hosting server. Just skip the middleman.
Shameless plug: If you want the data in dropbox encrypted, try https://github.com/lucas-clemente/git-cr :) I think I've also solved most of the concurrency problems, but haven't tested it yet.
bare git on dropbox is not dangerous in a multi user setup. I tested it a few years ago, the conflicts are minor annoyances but never destructive nor silent.
It's not, you get a conflict, so someone loses their last ref manipulation and you have to clear the conflict and update the ref again. You have a backup of the repo anyway as you cloned from it. So basically someone has to push again after the repo breaks (this is a bare repo in Dropbox with a local clone setup)
Right. And if that happens twice a year and it costs someone 10 minutes to fix, you could've just paid $12/yr to AWS CodeCommit and made money on the deal. Why would you half-ass such a critical piece of infrastructure when there are free and super inexpensive solutions that work properly?
I would not dream of doing it in a commercial environment. I do it in things like hackathons when you might be working with a random group, no-one wants to pony up to host it privately indefinitely, and messing around with permissions on AWS and registering everyone is a pain in the ass.
Share folder, git init --bare, done. Most people have Dropbox so its super easy.
That's what I use for my own projects. If I was working with other people I'd probably just setup gitlab (if not using github), with VPS providers these days setting something like that up is marginally more expensive than free.
Hmmm. I've been storing Git repos in Dropbox for... ages. Haven't tried sharing them, but for a single user I see no need to have an additional piece of software.
I am, however, quite interested in the potential for business-related repos - I guess it all depends on the authentication and transport mechanisms (I assume this will be using Dropbox's auth and TLS API)?
Have you used DB for your repos where you work on different machines that are not always on, and therefore there is a sync delay occasionally? And then get conflicted syncs on DB's end as a result when you try to interact with the repo before syncing is done across all machines? There are many headaches to just putting a git repo in DB and forgetting about it -- the one possible exception if you are only working on one machine all the time, with zero collaborators.
I use DB across many machines, some "permanently" on, two laptops and a (sporadically on) home desktop. Never had a conflict I couldn't recover from - and the 5-10 minutes some of the machines take to finish syncing are usually best spent catching up on e-mail rather than coding.
I've been doing this since Dropbox 1.0. Haven't had any serious issues since 2.x, and the benefits of having my complete working state available everywhere far outweigh the risks.
You may be ignoring the risk that people hear you're doing this and decide not to take you seriously as an engineer. There are clever Dropbox productivity hacks. This is not one of them.
If people were to form such an opinion of me based on that, I wouldn't take them seriously as peers.
There is no magic involved, and if a git repo ever breaks (which, ironically, last happened on a SCM server), I know enough about the internals to fix it by hand.
(although I would probably simply rewind Time Machine or pull from the central repo if it happened on my Mac - less time and hassle)
I just realized that by "DB" you meant "DropBox." I was reading it as "database" and thought you were putting MySQL files on DropBox! Hopefully my strong words make more sense now... apologies!
As anishathalye said you probably don't want to keep your git working directory on Dropbox. If you want to share a git working directory between machines, you can use a shared filesystem such as our ObjectiveFS (https://objectivefs.com). Note that if you use git with an OSXFUSE filesystem you currently need to add the "--quiet" option to work around an issue with signals (https://github.com/osxfuse/osxfuse/issues/213).
This is a remote helper meant to interact with a git repo hosted on Dropbox - it runs entirely on the client side - there's no special git-specific code that the Dropbox servers run.
Correct, you're not expected to leave the working directory in a Dropbox folder. In fact, that's a bad idea (for the reasons explained in the blog post / readme).
It's also recommended to selective sync out the repository directory so the desktop client doesn't touch the files at all.
My point being, if there is such demand from your customers for a particular feature that they hack it together themselves, or endure what commenters have expressed to be "tolerable strange behavior", then you should probably think about offering a solution.
Is it right to host your git repositories in a Dropbox folder? Probably not. Is it their fault? No. Is there sufficient evidence that quite a few people are trying to do it? Yes: https://www.google.com/search?q=dropbox+git
Unless, of course, they've become too big to care.
