#190Feb 13, 2016 · 40 min
Following the successful experiment of History of Hadoop, we are doing another Saturday experiment: an editorial podcast. Let us know your thoughts via Slack, Twitter, or email!
Our podcast errs on the side of technical rigor.
Whether the topic is distributed databases, microservices, Soylent, Uber, or Dwarf Fortress, we try to separate hype from substance, deferring the narrative to the guest. With that deference there is editorial objectivity on the part of Software Engineering Daily.
One Slack channel member said “when Software Engineering Daily’s opinions are announced, that takes time away from the guest.”
Fair enough.
However, as with any journalistic organization, we have opinions. On SE Daily, we stream objectivity and batch subjectivity.
This episode is willfully subjective. It has no guest. It is a monologue editorial inspired by Developer Tea.
Immutable laws are rare in software engineering, and when an engineer claims to have found one, that engineer is usually regarded with skepticism.
General principles are more welcome.
In this post and podcast episode, I convey some loose philosophies about modern software engineering. These are strong opinions weakly held. I welcome debate and discussion.
Software is a new field and nobody knows how to do it. If someone says you are unqualified and therefore you must do maintenance work, you should question that person. We have an upside down system where the people who are paid the least do the crappiest work. They tend to be young and naive.
This is not an axiom.
The narrative that is sold to young engineers by giant companies is the following: take your $80k/yr job, do the software maintenance which makes the company $1 million, and hate your life.
After you have spent enough time in the first tier of the intellectual strip mine, we will make you an SDE 2, where you can do slightly higher level refactoring for $150k a year, which will make the giant company $5 million. This is what is called an arbitrage.
We have an assembly line mindset left over from the industrial age. Software is more of an artisanship. Don’t believe that you are a replaceable cog. Don’t believe the one-size-fits-all interview process with whiteboarding problems. These serve to grind away your individuality and make you feel like an assembly line worker.
The planning and design process is an art, but once the requirements are in place you can proceed more deterministically. The same is true for the other quantitative activities I have taken part in–poker, music, and writing. As Michael Rosenthal and I discussed, the question of art vs. science is the same question as strategy vs. tactics.
I learned this very early on playing poker when I had to leave that career, and I had tightly coupled poker to my identity. If you make your job the same thing as who you are, then your self-worth is defined by those who are judging you in your job.
Your job is a means to service your own higher purpose.
When you take an action on your smartphone, there is latency before that action is ingested.
Servers sometimes will lie to you, but servers tend towards eventual consistency. The world works the same way. In the short term, human systems lie to us all the time, but the world tends toward eventual consistency–the truth eventually presents itself.
A decent analogy is the efficient market hypothesis: slowly efficient markets are an eventually consistent process.
The world is a distributed system–what is the consequence of this? We have to do the arduous risk and reward calculations that are mandatory for every distributed systems programmer.
Long-tail failures can and do occur.
In a distributed system, we often prioritize safety over liveness. In a distributed operating system, the programmer takes all precaution to avoid data loss. Similarly, if a real life scenario presents a small probability of giant downside risk, you should take huge precaution. If someone offers you to roll a die with 1000...
