Keynote, LambdaConf 2014 - The Silent Productivity Killer

Post on 06-May-2015

2.752 views 1 download

description

My take on what's wrong with the way we program - a sliver of what's wrong anyway.

Transcript of Keynote, LambdaConf 2014 - The Silent Productivity Killer

The Silent Productivity

Killer

Paul Phillips paulp@improving.org

Source: It's xkcd, yes.

Ex-Hacker Tells All!• One-time scalac hacker

• Now full-time malcontent

• Voted Mr. Congeniality by scala community

• Zero years in a row (and counting)

Jon Pretty admits it took him 10 years to amass 10 years scala experience.

The picture suggests he was enjoying the forty weeks vacation standard in Western Europe.

I dub thee Jon Pretty Slow.

< 10 years

Scientifically Determined Factsabout me, in the domain of interpersonal orientation

How Software Is Made

Initiate build

sequence!

TIME PASSES

You're older than you've ever been and now you're even older and now you're even older and now you're even older

!You're older than you've ever been

and now you're even older and now you're older still

!-- They Might Be Giants

Hmm, let's try it with that one brick three

millimeters to the left.

Wait a Minute• The titular "silent killer" is not the

time spent waiting

• It is the inhibition of flow state

• Outside of flow, you are but a shadow of your best self

Mihaly Csikszentmihalyi (mee-hy cheek-sent-mə-hy-ee) identifies these preconditions for flow: !1) Goals are clear 2) Feedback is immediate 3) A balance between opportunity and capacity

The Lag Analogy• Imagine editing system files on a faraway server

under a 6 second echo lag. No mistakes!

• Is the time to completion only 6 seconds greater?

• Is your error rate only "6 seconds" higher?

• Is the task only "6 seconds" less enjoyable?

• How often do we tolerate a "6 second lag" when writing code?

• To a first approximation: always

• Optimistically: way too often

The Lag Analogy

The Lag Analogy• "You should have been using screen/tmux!"

• Right: why tolerate such uncivilized conditions?

• A highly responsive layer coalesces inputs and batches them to the high-latency system

• It's madness to have to traverse every requirement to connect cause to effect

What IS a compiler?• Responsiveness is a (or THE) key to a UX

• It is mandatory to achieve flow state

• We are unwise to turn this critical task over to a general purpose compiler

• More responsive does not mean faster: usually it means slower! (i.e. less throughput)

repl startup• Initialize the compiler in a separate thread

• Print the prompt immediately

• Result: startup feels impossibly faster!

• Yet "minimum time to first result" is unchanged

• Responsiveness, where perception IS reality

multitasking• scalac is so slow to build, I always had two work

trees going simultaneously

• Context switching in that codebase is not cheap for the wetware

• Incremental compilation is useless in scalac

Interrupting Myself• Everyone seems to agree interruptions are costly

• I experience no interruptions - from humans

• Flow state is no less elusive for that

• "Flow interruptions" are everywhere

"incremental comp..."• Incremental compilation recompiles less (at best)

• I say: recompile never! Recompiling is rebooting

• First requirement is a functional compiler

• And we derive the output from the delta

A change to a base trait may incur the

recompilation of over one thousand files.

!The regeneration of 100

MB of class files. !

While the ΔOutput requires writing 4 bytes to

a single classfile.

Skeptical Hippo

"that's impossib..."• Yes, very likely it's impossible for scala, and most

existing languages

• But a language can be engineered around it

• And a language SHOULD be engineered around it

"who needs a comp..."• Bad news: most software cannot be reasoned about

• Try some pure reasoning against a few scalac bugs

• We can design better software, but that won't unwrite the rest of it

• And regardless, frobbing knobs and seeing what happens is how most of the humans learn

"The next step is to transfer some psychic energy each day from tasks that we don’t like doing, or from passive leisure, into something we never did before, or something we enjoy doing but don’t do often enough because it seems too much trouble. !There are literally millions of potentially interesting things in the world to see, to do, to learn about. But they don’t become actually interesting until we devote attention to them."

Mihaly Csikszentmihalyi