When a distinguished but elderly scientist states that something is possible, he is almost certainly right. When he states that something is impossible, he is very probably wrong.
The only way of discovering the limits of the possible is to venture a little way past them into the impossible.
Arthur C. Clarke’s third law, Any sufficiently advanced technology is indistinguishable from magic gets considerably more usage than the first two, which have never really entered the popular-collective unconscious.
The need for great performance really boils down to this one fact: You’ll never get create anything that feels like magic if your app is clunky and feels shitty. And if you can’t create magic, you won’t sell as well.
But if you create magic and your competition does not, you’ll beat them at the market nine times out of ten. People love magic, even when it isn’t-quite-magic.1
The fact that Clarke’s first two laws aren’t nearly as popular as the third isn’t much of a surprise. They aren’t glamormous. They aren’t prescriptive. They don’t really describe what “impossible” is.
The lack of popularity is not surprising, but it is a shame, because the first two laws drive us very clearly toward the third. When you push the boundaries of possibility, even in comparison to other experiences, you get delight. And if you string together enough delight, you approximate magic.
I think it’s important to look for solutions to performance problems in plain sight. I think it’s even more important to look for solutions to performance problems in plain sight when the surface area of the problem is large and/or unknown.
Beautiful and elaborate optimizations can be useful, but if they are taxing to produce, then they may be difficult to maintain, and if they are difficult to maintain — especially if they challenge the mental model of how the system works — then their usefulness should be carefully weighed against the costs that must be incurred to use them.
If the associated costs are high and the optimization makes the mental model more challenging, the optimization should not be used…unless the mental model was ineffective. Either way, something should be thrown away.
- First rule of optimization - Don’t
- Second rule of optimization - Don’t Yet (for experts only)2
When the US Gov’t found Ross Ulbricht, widely believed to be The Dread Pirate Roberts, the lead didn’t come from an anonymous tipster or a sophisticated and illegal intelligence monitoring technique. Rather, it came from rather low-tech information gathering and observation from an IRS employee who was brought in to support FBI and DEA agents who were having trouble finding meaningful leads.3
His keen attention to details that any other human being could have observed in the landscape the authorities had assembled thus far led them directly to Ulbricht.
The investigator was given the following quotation on a plaque:
“The world is full of obvious things which nobody by chance ever observes.”
This quote was penned not by Arthur C. Clarke, but by Arthur Conan Doyle, another acclaimed British author. The quote is attributed to his legendary problem-solving character Sherlock Holmes.
What approach would Sherlock Holmes take to solving performance problems on the web?
Per rule three, magic-that-isn’t-quite-magic may as well be magic.↩︎