Next time you get the urge to try out some new tech in production please remember this: 5 years from now, long after you got bored with it, some poor sap is going to be stuck fixing bugs in that new tech code you committed, because no one ever took the time to rip it out, and something in production depends on it. You’ve just introduce tech debt that will most likely be wasting someone’s time for many years to come.
I’m not saying never try new tech, I’m just saying, pay for it: write the new feature in the old tech FIRST, to keep yourself honest, then in the new tech, and deploy and maintain each of them on 50% of the servers. If the old tech wins, you up the capacity on the old tech servers, turn off the new ones, rip out the new code, and you’ve got no tech debt–experiment paid for in full. If you don’t do that you’re Trojan horsing it in–it sounds innocent enough to just upgrade one service to try out the new tech, but then what?
If the new tech ‘wins’, hold on a second and make sure you’ve thought through the implications…if the new tech wins does that mean you’re committed to porting 10M lines of code in the old tech to the new tech? What will that cost? What benefits will you get? Or are you instead committing to slowly port things over taking 10 years while maintaining +1 stacks increasing the cost of new feature development, maintenance, infrastructure, recruiting, training, bug fixes, troubleshooting, etc. Is it worth it? Really? $10M invested across 10 years in porting and +1 stack costs to achieve a 10% productivity boost once you’re done? If the math doesn’t work, maybe its not a good idea. And maybe do the math before starting the project.