skip to content W3C SVG

Choose Boring Technology

2017-11-09
  • Dan McKinley
    • 2015-07-23
    • Mature Engineering practices spoil you
    • How do you choose technolgoy?
    • How do you make developers happy?
      • “chase shiny objects […] is not really the [answer] at the highest levels of fulfillment. They don’t talk about nodejs in heaven.”
    • As a dev, Dan is now more interested in ‘the business track’
      • this is less about getting old and cranky, and more about moving up Maslow’s hierarchy of needs
    • the “innovation tokens” model
      • each company gets a more or less fixed number of innovation tokens early on in its life
      • you can spend them on any of your complicated, gigantic problems
      • do you want to spend these on your company’s core mission, core values, or on its technologies?
    • “There are known unknowns, and there are unknown unknowns”
      • there are classes of problems you can expect in a domain almost always
      • there are also consequences of your technologies' implementations that might shock you
      • Boring tech has fewer known unknowns and fewer unknown unknowns than new shiny tech
      • The failure modes of boring tech are well known
    • Optimize the total span and volume covered by your tech choices
      • “beating the polyglot programmer drums”
      • “using the best tool for the job”
      • implicitly, this is saying “the short-term benefits of using a new database exceed the long-term costs of operating it” in perpetuity.
    • Model problems and technology as a bipartite graph
      • every edge is solving a problem with some technology
      • every edge is a choice that carries benefit (having solved that problem)
      • every edge also carries cost or benefit (degree to which that tech is fit-to-purpose for that problem)
      • every technology node which is on an edge contributes cost (operating cost)
        • but you only pay it once regardless of how many edges
      • you want to minimize this: total cost = Sum(tech cost) - Sum(velocity benefit)
      • if the cost per tech is cheap, you can choose many; if it’s high, choose a few
    • In reality, tech choices are expensive
      • “Polyglot programming is not the kind of freedom we are looking for”
      • “It’s handing developers the chains so that they’re free to imprison themselves with operaitonal toil, forever.”
    • Example: activity feeds at etsy
      • built using suboptimal tools from the common stack (memcached)
      • memcached presented more upfront developer cost than if they’d adopted redis
      • however, over a few years the feature’s usage grew by 20x
      • because memcache is on the common stack
        • it probably would have been scaled anyway
        • it already had knowledgeable people on staff who would do this
        • the tech you depend on is neither abandonware nor “the thing you introduced that makes someone hate you forever”
    • The techs your company uses have broad importance: talk to each other
      • ask how you would solve the problem without adding anything new
      • if solving a problem with existing techs requires resorting to unnatural acts, write down those unnatural acts
      • if you try a new tech, find low-risk ways to get started
      • commit to removing the old, redundant system
    • every piece of software has a curve of “time since you started using it” vs “kinks you haven’t worked out yet”
      • there’s a natural tendency to abandon tech early in the curve, before mastery/stabilization