[Pacabel]

While it may be perfectly acceptable to ignore the lower-level details when first learning some new technology, or even when using it for personal use only, that sort of ignorance is not acceptable when working on systems that are meant to be used seriously.

At that point, you need to have at least a high-level understanding of each and every layer that you're building upon. If you don't, then it will come back to haunt you, whoever uses your software, and whoever has to maintain it after you're long gone.

We've all encountered, or at least heard of, the Java and Ruby programmers who use ORMs, but who refuse to learn SQL, and refuse to learn how relational databases actually work. They write their software while understanding only the highest levels of the abstractions that they're working with, and then the system ends up performing horribly. Then somebody more experienced and knowledgeable gets called in, and gets to see how the ORM is generating abysmal SQL queries, or indexes aren't being used.

It's better to know and understand than to not, especially in the long term.

[bguthrie]

I have built many systems that were meant to be used seriously. Never once has my lack of understanding of graphics cards, for example, come back to bite me.

Every system is built on abstractions. You can argue about the value of understanding particular systems, but we are well past the point when any one person can or should understand them all. No one is arguing that it's acceptable to learn only the highest-level abstractions. But it is acceptable to let somebody else worry about some of the lower-level ones.

People who don't focus on understanding engineering systems often focus on understanding other systems––team dynamics, for example, or business needs. I've met a lot of developers who only cared about learning the technical side, and were happy to look down on developers who didn't feel the same. Is that wiser, somehow?

[Pacabel]

What kind of systems were you working on?

No, you probably don't need to know much about how graphics cards work if you're working with databases or other software that doesn't really involve graphics in any serious way. But if you're working on graphics card drivers, high-end video games, visualization systems, or simulations, then you'd probably should have such knowledge.

The point is not to fully understand every single layer that exists, including those that your software never interacts with, or only just makes very light usage of. But you should have at least a basic understanding all the way down through the code and hardware you do use, directly or indirectly.

[ampersandy]

It's perfectly acceptable to not understand the low level details of seriously used systems. Isn't that what teams are for? You can trust the engineers developing infrastructure so that you don't have to worry about it. Sure, understanding the high level concepts can't hurt, but it isn't going to make a front-end developer any more productive.

[weland]

> It's perfectly acceptable to not understand the low level details of seriously used systems.

Yes, that's what my colleagues thought a few months ago, too, when they churned up a web UI for an embedded device. My teamwork was, apparently, insufficient to make them realize that a configuration interface that takes half the available flash space is rather large (the other half was taken by a goddamn kernel, a web server, a proxy server, a bunch of other networking tools, a logging service and all the tools that communicated with the web ui!). Lack of understanding regarding how routing works in TCP/IP networks also meant that their first attempt had a pretty fatal flaw which resulted in the configuration interface becoming inaccessible, which became apparent in the first ten minutes of testing and required a significant amount of redesign because the entire workflow of the interface depended on that.

Their Bootstrap wizardry was remarkable, but ignoring how the underlying system worked resulted in them having to start from scratch.

It works the other way 'round, too. Not understanding how web tools work would also make my systems programming entirely useless when it has to provide the infrastructure on which such tools rely.

No one likes all this complexity. I hate it, too. I sometimes truly, really wish my job was as simple as taking this thing from here and putting it over there. Ignoring this complexity doesn't make it go away though; building tools that can isolate some of it behind a comprehensive and consistent interface is ok, but relying on those tools to do it for you without understanding what they do is technical suicide. It's this kind of thinking that makes people think it's ok for a rudimentary word processor to take six fucking seconds to load, after asking you for a password, in 2014, on a quad-core system.

Edit: it's also worth pointing out that this is really beyond the scope of the article's argument. What the article insists is that this:

    var numbers = [7, 8, 1, 3, 2, 7, 6];
    numbers.sort();

hardly helps anyone learn coding as a problem solving tool. This exercise taught the reader exactly nothing about how to solve the problem of sorting numbers. I could ask someone who works for me to sort me those numbers, and my understanding about algorithmic thinking would be as rudimentary as before.

Teaching children how to use calculators does not increase their problem-solving ability, not by an iota. Teaching them mathematics does that, and their ability is further advanced when they learn about theorems and structuring proofs, about induction and, yes, about using a calculator.

It's the same here. I'm not arguing that computer science should be purist to the extent that you shouldn't use anything but a Number 2 pencil to program. That's stupid. I am arguing, though, that teaching people "to code", as if programming consisted in nothing but translating from English to Codese, is pretty much as retarded as teaching mathematics as if it were a language, consisting of nothing other than translating the English "Two apples and three apples make how many apples?" into the far more cryptic "2 + 3 = ?".

[dhoulb]

Teaching kids to use calculators DOES increase their problem solving ability. They can produce the same result as someone using paper. If the problem's really advanced, they may have to use brackets correctly too. You teach the same logic whether it's on paper or not, just instead of shuffling numbers in your head, a system that's BETTER at it is doing it instead.

But the issue is, you've added the Calculator dependency to the Kid program. Whether or not that's an issue depends on whether Calculator is a well maintained and dependable project.

On the other hand, there are downsides to NOT using the easier Calculator library, and writing all the code from scratch (i.e. Memorising Mathematics systems and practicing them enough to use them at any reasonable speed) — 1) storage space and using up brainpower / focus storing micro numbers, when you could be focusing on the bigger picture algorithm, 2) speed, the Calculator library is optimised and written by experts, 3) bugs, the Calculator library is well tested, whereas it's easy for anyone to make simple arithmetic mistakes.

[conradev]

I disagree with the last point. Understanding the low level details of a system definitely makes a developer more productive.

When I am working on an iOS app and I get stuck on a bug, I can open a disassembler and immediately figure out the cause of the bug. The same goes for a ruby developer who wonders why their queries are taking forever or a front end developer who doesn't know why their interface is dropping frames.

Any serious engineer should be able to fix these problems on the spot if tasked with it.

[Rusky]

The engineers can't really do that for all cases [1]. Understanding abstractly what's going on at least a few layers up and down your stack will definitely make a front-end developer more productive, if by productive you mean able to complete a finished product rather than just a buggy first draft.

[1] http://www.joelonsoftware.com/articles/LeakyAbstractions.htm...

[hateCS]

> While it may be perfectly acceptable to ignore the lower-level details when first learning some new technology, or even when using it for personal use only, that sort of ignorance is not acceptable when working on systems that are meant to be used seriously.

I'm a self-taught web developer. While I have probably picked up some programming fundamentals/best practices through experience I have very little interest in computer science or learning the low-level details of every technology I use.

I probably couldn't get a job at most SV/SF startups, but I run several WordPress sites that have millions of monthly visitors and that have traffic spikes of 100,000+ visits in 2-3 hours at least a few times each month. I'm paid very well and I don't need to apply rocket science to do a good job.

Meanwhile I have friends with CS degrees at hot VC backed startups using all sorts of crazy technologies and advanced programming techniques for sites that get maybe 100,000 visits a month.

If you ask me, ignorance is premature optimization, overengineering and hiring expensive people to apply skills you're nowhere close to needing.

[Pacabel]

I think you've benefited very heavily (a lot more than you may realize) from the work of people who did have more formal training, and who used that knowledge to build the systems that you in turn have built upon.

You're right, it does take much less effort and skill to run moderately high-traffic WordPress-based web site these days. That wasn't always true, though. In the 1990s and even into the early 2000s, it did take skill to get the most out of web servers with clockrates below 200 MHz, maybe 16 MB of RAM if lucky, very limited storage, and limited network connectivity. Specialists were needed in cases like that.

What you seem to consider "ignorance" is usually just considered "forethought" and "planning ahead" by those involved. It's easy to rail against so-called "premature optimization" until you've seen that kind of prudence prevent unexpected scalability disasters from happening. For a young company, like the kind your friends apparently work at, a little bit of foresight and effort at the start can significantly mitigate the harm that can come from unexpectedly large demand. In some cases this is what will determine whether or not that company survives and thrives.

[hateCS2]

> I think you've benefited very heavily (a lot more than you may realize) from the work of people who did have more formal training, and who used that knowledge to build the systems that you in turn have built upon.

And your point is? Am I supposed to create a shrine to the people who built the first microprocessor?

> You're right, it does take much less effort and skill to run moderately high-traffic WordPress-based web site these days. That wasn't always true, though. In the 1990s and even into the early 2000s, it did take skill to get the most out of web servers with clockrates below 200 MHz, maybe 16 MB of RAM if lucky, very limited storage, and limited network connectivity. Specialists were needed in cases like that.

I'm sorry but it's not 1999. I can't imagine anyone winning a potential client over today by saying "You couldn't do in 1999 what you're doing without someone like me" or "I will architect your website so that it can run smoothly on a server with 32MB of RAM."

> What you seem to consider "ignorance" is usually just considered "forethought" and "planning ahead" by those involved.

So struggling to hire (because you're looking for overqualified candidates a lot of whom don't want to work on a simple website with less than 100,000 visits per month) and increasing your burn rate by paying $140,000/year fully loaded for web developers you can call "engineers" is "forethought" and "planning ahead"?

I know enough people who have worked at typical VC backed startups to know that most of them never grow big enough to have the kind of scalability and performance challenges they'd all like to believe they're going to have. And as a practical matter, if you look at the great companies that have emerged in the past decade (think Facebook, Twitter, etc.) and how their systems have evolved as they scaled, it's pretty clear that if you become a massive success, you're going to need to substantially rewrite your code base anyway and by that time you won't have to hunt for the highly technical engineers you need. They will be knocking on your door.

The biggest risk to just about every new startup is not having a developer who is a little too friendly with an ORM and who might have to turn to StackOverflow for help with a complex SQL query. It's getting a good enough product launched before you miss your market window and/or run out of cash.