A Technologist Who Speaks Business


6 years ago
Source Control – Calling for Backup

This second in my Calling For Backup series is a quick argument about why source control is not optional, period.

The Scene

You’re working for a start-up in your parent’s bedroom. Development is cruising at a million miles an hour fueled by Dr. Pepper and swagger. Or you’re on a small team inside a large corporation–maybe doing skunkworks development, or maybe working on a thankless legacy application that only you (at best) know how to understand. Maybe there are one or two other developers, but largely everyone knows everyone else and the state of the code can be held in one person’s head.

Nobody on your (small) team wants to take the time to implement source control because they think it’s “too complicated” or “a hassle” or “will take too much time”. You missed the easiest chance to do it at the very beginning of your project when there was no code, so trying to do it now involves reconciling the local copies everyone has on their PCs or shared drives with everyone else’s copies.

Or maybe you do have source control, but it’s anyone’s guess is what’s deployed in production is actually the same as what’s in source control; so for all practical purposes you really don’t have source control. Again, the only way to fix the problem is to painstakingly reconcile production with source control.

“Isn’t it easier to just deploy what we know has changed? It’s worked fine so far and we have so much other ‘real work’ to do that will pay the bills,” says your boss/investor/co-hacker.

The Real Problem

Your team is suffering from a bad case of short-term-ism and poor risk estimation.

Allow yourself to imagine for a second if your primary production machine and/or your primary development machine went up in smoke. What would you have lost? How long would it take you to recover from the loss? Could you even ever be sure you did recover everything, or might you just have to cross your fingers and hope you remembered it all? All of these questions are measures of risk–and its consequences.

But the obscene risk you’re running isn’t the only problem. Try honestly reflecting back on how many decisions you’ve deferred because you can’t be sure you can undo them. Is there a big refactoring you’ve been wanting to do, but are afraid to because it’s too much change to handle all at once (what if you change your mind?). How about a new technology or library you’d like to try out, but haven’t because you can’t be sure it will work out and don’t want to manually roll it back if it sucks? These are hidden costs to not having source control.

Lastly, you are losing out on productivity gains. Imagine how much cheaper it would be to outsource your commodity development to Brazil or Ukraine, if only you wouldn’t have to give those outsourced developers access to your production hardware just to change a stylesheet or two.

Why It’s Hard to Explain

There are three reasons why source control can be a hard sell:
1) it’s surprisingly difficult for people who haven’t used it before to understand how it works
2) many of its benefits revolve around “what ifs” and “if onlys”. Your detractors will demand specific examples of what might go wrong, but by definition you can’t predict the unpredictable.
3) all of its costs are concrete and immediate, but its benefits are intangible and/or long-term.

The Rationale

Not using source control borders on professional negligence. It’s item #1 on the “Joel Test”. I doubt you’d even get beyond level 0 (hint: that level doesn’t really exist, it’s so low) in the CMMI maturity model without source control. Steve McConnell Himself would slap your grandmother if he found you without source control. But that’s all just assertion by authority. How about some concrete reasons…

If you are a software company, then the only real asset you have is your source code. Not having your source code in source control is equivalent to Walmart not locking the doors on its stores when they close: it’s just inviting a catastrophic loss. Even if you aren’t a software company per se, it’s likely software is a significant part of either your product or of the way you make your product. Without it, you would be as hosed as a retailer if the mob had robbed all its warehouses.

In addition to protecting you from loss, source control allows you to roll back. The benefits of rolling back aren’t as obvious at first, but they include:
1) Allowing you to do “spikes” to test out new ideas, without having to worry about backing them out if you change your mind
2) Helping you roll back individual features just before a release if testing shows them to be a problem
3) Allowing you to upgrade libraries without worry that you can’t go back if there are problems
4) Undoing crappy work the a new/untalented/sloppy developer has done

Lastly, source control offers you a convenient way to share code. Instead of zipping up your code and emailing it to someone, you can just point them at a URL. This is especially useful when you want to outsource development, or when you have a distributed team.

If you are having a hard time getting your team to understand the concept of source control, then Eric Sink wrote a good guide to understanding source control. Disclaimer: Eric Sink’s company sells a source control product. The nice people at Subversion also have a good primer on source control. There’s also a good MSDN article on branching and merging

Some Real Stories

My best “catastrophic loss” story is from a web development client. When I got to them, the client wasn’t using source control. In fact, they were doing development either directly on their production hardware or in Dreamweaver on their own Mac and then pushing the entire local image out to production using Dreamweaver’s FTP function.

When I came on I basically refused to work this way. I created my own Subversion repository and put all my work in there. When I needed to release, I copied the files I’d changed (which Subversion helpfully highlighed for me) to their production server by hand. This all worked reasonably well, until the client hired a second consultant in addition to me. That second consultant refused to use source control, insisting that it was “too complicated.” So from time to time we would step on each other’s changes and basically cause problems for each other. During these times, entire parts of the client’s site were broken. No amount of begging could persuade him to change.

Finally, about 6 months after I left the project, I got an email from the client. She had accidentally pushed out a bad version of the site from Dreamweaver and had basically hosed the entire site. Luckily, I was able to pull my latest version out of source control and have it restored for her in 20 minutes. But any changes that other consultant had made in the last 6 months?…..gone.

More recently, I was working on a team that had to push out a big release just after one of their senior developers had quit. When they entered user acceptance testing, they discovered that the very last thing that developer had checked in before quitting was causing lots of problems and needed to be taken out. Unfortunately, the changes were shotgunned all over the code base in tiny chunks. Luckily, they had source control to understand what he had changed and to sort out the mess and were still able to release on time. If they hadn’t had source control the only choice they would have had was simply barreling forward into the unknown.

Guerrilla Work Arounds

If you simply can’t convince your team to use source control, then at the very least do what I did and create a repository for yourself. It will only take one time when that repository saves the day (and it will, believe me) to convince the rest of the team to switch over. Both Subversion and CVS are free and will install on your local machine if necessary. But that isn’t even necessary because there are plenty of free or low cost hosting services, too.

♦ End

Comments are closed.


Run a SaaS app and want to know how to improve customer retention?

Sign up for our free 5-week email course on combating churn.