Ask HN: How would you convince your company to start using version control?

Didn't Joel write an article about that sometime ago? Ah, here it is: http://www.joelonsoftware.com/articles/fog0000000332.html The point of it is, use it yourself. If you use subversion or whatever else (seems you are familiar with subversion so go ahead and use that) then you can roll back changes, you can do diffs, you have a full copy of the source with revisions, and you can always tell rsync to ignore the .svn directories when you copy things over to production. Eventually something will break, and you will be the guy/girl with the magical powers to bring it back to a non-broken version and find what has changed. Eventually, it will trickle to the rest of the team that it's not a bad idea. Version control can be incrementally adopted. Just sneakily do a commit after other people's changes, and show people how it works. If not everyone adopts it, fine. There is certainly a strategy that's necessary to make sure you don't step on each other's toes while editing things. Use that to mediate between the source control crowd and the non-source-control crowd. Forcing it on others is likely to be counterproductive. Start using it yourself and let others see what happens.

I just wonder that if they're not savvy enough to use a RCS what other types of battles will you have to fight in the future?

If I was in your shoes, I'd start looking for another job with a company with a bit more technical sophistication. Somewhere that the battles you fight aren't over trivial technology problems and practices and issues that already have solutions.

Now that doesn't solve your problem, so probably the best thing is start using an RCS yourself, then create a build system that will push changes out to the server. You could initially sell this as a time-saver in updating the server. Then start selling the whole history part of using an RCS to go back to previous versions of a file.

But, if they're not using an RCS in the 21st century, what else are they doing wrong, and should you really stay there? :)

I am having this same problem with our development team (mostly c) at my job. They find subversion too arduous but then don't understand why there are so many problems and bugs in their releases.

Thank you for mimicking my sentiments precisely. I want to fight battles that aren't over trivial technology problems that already have solutions!

Start using it yourself, maybe even with a cron job that regularly pulls the latest changes into your local repositories. I give it 2 or 3 version conflict or deleted file incidents where your use of VC saves the day.

I feel that there may be a chance of "Why would we need to user version control when we can just use you instead?". I do like the idea though and will probably set up a repository today for projects that I'm working on.

That would be you taking on more responsibility? Reason for a raise if you ask me.

When I started my last job they weren't using version control, so I just told them that if nothing else it would be very easy and painless to just check in every time we push to production so that we always have a snapshot of the last version. The problem with them was that their developer would go and make changes to files in production to fix bugs and issues, even after started using version control, so it was always a chore keeping up with his hotshot prod changes. You don't have to implement a full continuous integration environment or start branching each developer separately, you can just start catching snapshots along the way and it will have a positive impact on future development.

The right solution to that is to overwrite his "hotshot prod changes" from version control until he learns that he has to do it in version control first.

If you don't have enough political pull to make that stick, then you have a second problem...

Or you check in the production changes....

If you can't do this from your source control system, try something more powerful.

That is a workaround but it has two major shortcomings. The first is that you aren't getting the other developers to use source control. The second is that you wind up with the challenge and the blame for breaking it in the case where someone fixed it in production, someone else checked a different fix into source control, and you are expected to merge them.

The simple rule, "You knew I was going to go from version control, your fault for not fixing it in there" is a proven technique for getting people to actually use version control. I've seen it done, and know people who did it at several other organizations and it works. (Though at this point my first impression of developers that don't know how to use version control is that they must be incompetent people I'd prefer not to work with. Not using version control a decade ago was somewhat more excusable than not know about it today.)

>"You knew I was going to go from version control, your fault for not fixing it in there" is a proven technique for getting people to actually use version control. I've seen it done, and know people who did it at several other organizations and it works.

I know being an ass to people works. Tried and true for sure, very passive aggressive technique to get developers to do things.

Also just keeping a production branch and development branch works too. Check in the production changes to production branch.

Production hotshot doesn't port over changes or at least mention the important ones to the guys working on the dev team? Kick that poor behavior to his boss about how he's not pulling his weight, staffing problems are HIS problem not yours, and by strongarming the guy with passive aggressive overwrites, you're screwing the company AND the hotshot, rather than just the hotshot.

I'll second that. It's politics, and unfortunately, you need a business case to push policy. Wait until the next production outage caused by nasty, ad-hoc changes and point at the version control system and say, "If we had this, we could just roll it back."

I'd almost say your build process has to accommodate version control first. Capistrano is pretty good at that.

If you have a developer with production access, you're bound to get stuff like this.

one advantage you might use to "sell" the idea is that it lets people work on laptops away from their desk more easily (i'm imagining that a marketing company is full of people wandering round, brainstorming in cafes, etc... :o)

I hadn't considered the usefulness of this for the designers. Being able to go pull up "that awesome design from last week" would probably be very attractive to them.

I'd say you should start using Git or Hg or any distributed system immediately. No need to worry about getting complete adoption for those to be exceedingly effective. Svn (or any "centralized only" SCM) is less capable in an environment such as this as it requires everyone to adopt at once.

After you have a week or two worth of changes in your repor then you can convert people one by one, and you don't even have to tell people about the repositories you're using until it's time to try to convert them.

I would tell him you want to setup a source control server (That you personally will administer so no overhead or extra work for him.) Have it be fully optional aka people who use it have to fully comply with current practices(Everyone has to be able to work as if it doesn't exist.) That way he gets zero complaints from other staffers who resist change.

Drop a note explaining potential disasters of using such a system to everyone including your lead developer's boss. Then wait for the disaster to happen. And then shout again pointing to your email.

Perhaps put a business case forward with the benefits of transferring to a version control system?

For example: * Easy rollbacks * Easily work on separate branches * etc