Hacker News Reader

Comment by apple314159 | Hacker News Reader

Not sure what value proposition is here. Static content is well understood. I'm using google storage and it does the job pretty well.

View on HN

It is not to store content but the ability to create new sites just by creating a new folder, e.g. if you create a new folder in 'Dropbox/Apps/websrvr' called 'apple314159' you would have a new website created and online at 'http://apple314159.websrvr.in/' and you could manage your website through your local computer.

Dropbox doesn't play well with git, or perhaps version control in general. Having a live site which is unversioned is somewhat less than desirable. I am sure there is some category of people for whom this is useful, but I am a professional, which means I have been burned too many times to consider not using version control. Hence, this service is entirely useless to me.

Hiding complexity from the user is mostly good; it's a shame that the replication service isn't more capable.

You could still use git to version your website and then websrvr to host your website. I do it for 'http://www.websrvr.in/', you can checkout (https://github.com/websrvr/www.websrvr.in/) for it's source code. I use middleman to build the app and use the "~/Dropbox/Apps/websrvr/wwww" folder as the target folder, it's works out nicely.

I keep about a dozen personal, non-critical projects in Dropbox. All are versioned with git, and it works great. Dropbox itself is versioned (since I'm a paying subscriber), so I have two layers of security there.

What negative experiences have you had with Dropbox + VCS?

Maybe it's better now, but I've had problems with .git folders in dropbox. Dropbox doesn't understand the directory structure and conflicts are handled by merging competing .git folders and renaming internal files as needed -- almost certainly not what you want.

Storing .git folders in dropbox doesn't seem like the best idea. You could keep a git repository in a different location and have a small build script like this https://github.com/minhajuddin/notes.minhajuddin.com/blob/ma... which builds your website and copies it to a dropbox folder. With this approach you could even use something like http://middlemanapp.com/.

Why? Security? Something else?

I have 2+ side projects or submodules under development at any given time. I've been storing them in Dropbox for the last 3 years. I've been working on them actively from 3 different machines (two Ubuntu desktops and one OS X laptop), and they're turned off -- so not syncing -- when I'm not using them.

All that said, I haven't had one single problem. So empirically speaking, I don't see any problem with the mechanics of .git folders in Dropbox.

(Note that my key files, API keys, etc. are not synced to Dropbox.)

I have had many instances of copies of files being created when there is a conflict. This shouldn't happen if the syncs happen before you work on a file (which is your situation). The main issue for me is that dropbox doesn't explicitly tell you that there is a conflict, it just creates a file with the other machine's name and you may end up scratching your head for a long time before figuring out about the conflict.

To be honest, I haven't tried storing git repos on Dropbox so these are just my assumptions based on dropbox usage for other files. I am just afraid that it may end up corrupting my git repo.

I can't speak for before I started using Dropbox with git, but for the past 3 years, conflicts are handled well (and not as you describe).

First of all, Dropbox will not overwrite data. If it finds a conflict, it will save a second file with "[machine name]'s Conflicted Copy" and the modification date at the end.

This hasn't happened to me with any git files yet, however, because my machines are always online. That means my working tree is always the same across devices and my .git folder is always identical.

The only way a conflict would be created is if I committed while working offline and then committed on another machine while working offline.

If that happens, I could just grep my Dropbox folder for files with "conflicted" in the name and remove them. Or, because I committed my files to a remote (often Github, but sometimes others), I could just delete the local copy and clone.

Sure, I'm not suggesting there would be permanent data loss... but doesn't Dropbox renaming a bunch of files deep within .git folder result in an unusable repository that you'd have to fix by hand or restore from backup?

This whole discussion made me curious, so I tested it. Dropbox doesn't rename files. It creates new files. All you have to do is delete the new files (which are really "forks" of your current data) and you have your working tree synced with your data store again. It's really easy.

Value prop is obvious: do you change your car's oil yourself or take it to the shop? Some DIYs will set this up on AWS while others will find it 'convenient' to spend $3/mo and be done with it.

Reply on news.ycombinator.com