Very nicely put. I mean look at a car: It's essentially a lot of 3rd party components put together coherently. I don't think any part of my Peugeot has a Peugeot label on it, rather Bosch, Autoliv, Michelin, etc.

Same thing with airplanes. Boeing is pretty much Honeywell+GE+Liebherr to name a few.

I am unsure why people find componentizing and integrating unrewarding in IT. I didn't hear my dad (a sound engineer) complain that he no longer spins a wire onto a metal to obtain a speaker. :)

Let me work with your analogy a bit. What if your dad's job was to install sound systems designed around poor quality speaker components, using expensive DSP systems that add a lot of latency to the signal and draw lots of power?

He knows he could build better components than the ones he uses if he put the time in, and create a better overall system design with a better price/performance by applying state of the art knowledge, but people prefer to pay him to combine commodity parts of questionable design in the standard, proven, but clearly suboptimal way.

Now where is his job satisfaction?

If you really wanted to stretch the analogy, maybe imagine the expensive DSPs he is expected to install all forward a record of any audio played on them to a shadowy government agency.

I imagine a lot of product engineers cringe when they are forced to make a design choice that will save 1c on the BoM but causes the device to fail sooner or be impossible to repair.

I understand your analogy, but I don't think it is that bad. Rather, instead of wiring one perfectly-sized speakers, a sound engineer might have to use two off-the-shelf speakers of a smaller size.

I understand the frustration of not doing things optimally from a theoretical point of view. However, the unoptimality is offset by using battle-tested, robust components.

One issue I've seen is that a person who devotes 100% of their time to integration can lose their quantitative engineering skill -- it's a "use it or lose it" phenomenon. The folks who manage to cling to that work end up becoming the only ones who can do it, resulting in a sort of caste system with no relation to actual productivity.

However, in my experience, the best integrators are worth their weight in gold. It requires a mental organization, patience, and discipline that I don't possess. The ones who do it badly are the ones who are racking up dependencies and technical debt.

O(m x n), where n is the number of pieces and m is the number of systems it needs to interconnect with. O(n^2) is the asymptotic upper bound, which is never reached, because a piece needn't be connected to itself: m < n for all n, but m/n can approach 1 as n goes to infinity.

It wouldn't be Hacker News without such ponderous pedantry!

I think so. That's why eventually you shatter your monolith into micro-services and then have 1000+ people work on coordinating how these services deal with each other.