OR Route 38 erosion, Jan. 2012 storm

In every company, there is a tendency to value code that was invented in-house over code that was, to put it bluntly, Not Invented Here. There is an eternal struggle to find balance between the convenience of pre-packaged code that is not fully vetted and the trustworthiness of code they themselves have written. As is typical in these tales, Jon's company got it wrong.

When Jon was asked what logging solution he was planning to implement in his company's .NET-based application, he gave the right answer: log4net.

No sooner had he spoken, however, than the room grew quiet. The locusts could be heard clearly from outside the window, screaming their disgust at the idea. Not-here! Not-here! Not-here!

"Or ... maybe I could throw something together?" he suggested timidly.

"Splendid!" replied Jon's boss, Ned, heartily. "You can use my logger as a head start. Rolled it by hand back in my university days. It's much better than that log4net crap. Did you realize that piece of dung uses reflection? Reflection! Something as simple as a logger doesn't need anything as expensive as that."

And Ned was right, in a way. His logger didn't use reflection; it hardcoded the method name into each log call made. But that's not all. In a bid to save on "expensive" new features Ned barely understood, the logger was held to a strict 2.0 code level, despite the project using the 4.5 framework. And in a flash of coursework-inspired brillance, he had implemented his own custom stack to store incoming logs—a move that entirely undid any cost savings achieved by avoiding reflection three times over.

But Jon tried. He rewrote the stack structure, desperate to reclaim some of the cycles. He documented methods that confused him, hoping to leave the codebase better than he found it. And he tried and tried to make it work consistently.

The days turned into weeks, and the project deadlines began to slip. Ned grew sterner and sterner with every depressing status meeting. "That logger can't possibly be responsible for all this! It was fine when I wrote it!"

"But sir, if we just pull it and use Log4Net—" Jon began.

"I don't want to hear another word about Log4net! Just get the project done!"

And so it was done—weeks behind schedule, and with a tendency to drop log files in the event of a crash, just when you'd want them the most. Jon privately swore he'd never touch the logger again; come hell or high water, he'd use log4net next time and be done with it.

It was three months later before he had another chance. Another desktop app needed to be built where logs would be mission-critical in the event of a disaster.

"Sir, about the logging—" Jon piped up.

"I know what you're going to say, and I won't hear of it," Ned cut him off.

"But sir, if you'll just listen a moment—"

"I said no and that's final: there's no way you'll be using that POS logger you put into production last time! No, you're going to have to write something custom from scratch, no getting around it!"

Jon put in his custom, from scratch, two weeks' notice that very day.

[Advertisement] BuildMaster allows you to create a self-service release management platform that allows different teams to manage their applications. Explore how!