All The al3x That's Fit to Print
I’m in the New York Times today. It feels weird.
I talked to the reporter over the phone three weeks ago. A week later, they decided to get the story ready for press and sent out a photographer. I happened to be at my father’s home in Northern Virginia on a trip back to the DC area. Hence the wallpaper.
I’m happy with my place in the story, but part of me wishes it was a project I was more passionate about. I made downforeveryoneorjustme.com for a friend on a whim, utterly unconnected to my work at Twitter. I’ve spent maybe six hours on the site, total, ever. Most of that was setting up the domain and hosting and doing the original Ruby implementation. I then spent a couple hours moving it over to Python for Google App Engine with the idea that I wouldn’t have to think about uptime. As the story mentions, that didn’t exactly work out, but it’s been better than futzing with a VPS or three.
It’s good that a conversation about the cultural and business implications of downtime is starting, or perhaps continuing and deepening. It’s certainly something I’ve been forcibly immersed over the last year and a half of helping to keep Twitter afloat. But it’s also a conversation I’m weary of.
Of late, I’ve tried as much as possible to focus my time at Twitter on building a new system that works at scale and does so predictably and measurably. That’s not easy when the current system is still on fire. So much of the last year of my professional life has been spent reaching for buckets of water that it’s hard to stop myself from trying to stomp out every new blaze. If I don’t, though, the new components that we hope will make the system work get pushed out further and further.
Figuring out how to work on version 2.0 while part of the team is maintaining the 1.x series is a classic software development management problem, and one that’s only become more complicated in the web application development space. The web offers incredible freedom to see if an idea will succeed in the marketplace. The trade-off is that once you have a successful idea, you only have so long to make it as dependable as death and taxes.
At Twitter, we’re well aware that we have essentially the most patient and forgiving set of users we could ever hope for. I don’t intend to test their patience a moment longer than we have to. In the meantime, seeing our site occasionally float to the top of the request patterns for downforeveryoneorjustme.com is a stark reminder of all the work we have left to do.