Writing software is easier in 2020 than it was when I started in 2002. Coding education is more accessible, Coding platforms are more inviting, languages are more forgiving, IDEs are more helpful, we have millions of answers on stackoverflow, and there are open-source libraries that have probably implemented whatever irks you. That being said, for the professional, I have found no substitute for the academic rigor of a good CS curriculum to shape how we think about organizing and engineering code well.

I originally posted this (in 2013! reposted today prompted by HN's second chance pool) because it struck me how, with very few modifications, this exact article could be republished today. I find it fascinating that we can be having the same arguments that people a half century ago were having, with little to no awareness that we're repeating the exact same things. It makes me realize that perhaps software is not as young a field as we like to sometimes pretend (it's common to read on HN that e.g. software is so young and immature compared to civil or electrical engineering, etc)

It's also interesting to dig into the author's name - apparently a half century ago he had some reputation in tech circles, but as far as I can tell he's mostly forgotten today.

https://www.theatlantic.com/technology/archive/2016/05/what-...

He sadly seems to have passed a couple years ago:

https://www.legacy.com/obituaries/name/erik-sandberg-diment-...

I like the bit at the end about studying the classics in Latin and Greek. It helps me see that "language" is not quite the right word for programming systems, and that this wrong word choice has led to writers falsely thinking that learning to program is like learning a second language.

But a language is tied to its execution context and semantics. This leads to either dividing up languages into "natlangs" and "conlangs" depending on usage patterns and style, or to studying programming solely from the systems perspective and ignoring linguistics altogether.

I wonder how things would have been different had we, as a community, rejected this terminology and stance. What if, even further, we had rejected the idea that computing can be made "simple" or "intuitive" or "mainstream", and instead forced folks to learn programming to APIs in order to even use computers.

Eloi don't need to learn programming: "[by] the time you became truly proficient at programming, chances are that whatever you set out to write would be available in some form from a software publisher."

Morlocks might want to learn programming, not because it's useful for eating Eloi, but because they're the sort of people for whom "purchasing an automobile for a cross-country trip [and] first [studying] cartography, then [proceeding] to obtain aerial and satellite photographs of the proposed route, and finally [drawing] a detailed map for the whole journey" sounds like a brilliant yak shave.

(Sandberg-Diment has left out the parts where obtaining the satellite photographs involves SDR hacking into downlink telemetry and drawing the detailed map first requires implementing a geometric-algebra based direct-to-framebuffer renderer)

That was fascinating!

My mother purchased my first computer (an Apple ][e from the local Macy's) in 1983. We didn't have much money to purchase software. I didn't even understand there was a software industry, let alone where I might purchase it.

But that computer came furnished with some basic software that allowed one to write BASIC and, also, to use a mouse with a paint program (yes, before Macintosh debuted).

While his main point that not everyone need, nor should, learn to program computers, what Sandberg-Diment misses is the sheer size of the burgeoning home computer market and how the personal computer would revolutionize and fundamentally alter the world.

Reading "Personal computers: does everyone need to learn programming?" is slightly shocking for me me because having lived in that world, the difference between what Sandberg-Diment casually suggests and its real-world manifestation could have never been forecast.

Two examples:

> First, it allows you to develop software that is not available commercially, and in some cases it lets you customize purchased software to serve your specific needs better.

The ability to modify software "to serve your specific needs better" is a general gesture to client-side scripting, software consulting, and even FOSS. Linux did not exist in 1984 (Torvalds was 15 years old and his magnum opus was still 6 years away). Empires can (and did) fit in the gap between Sandberg-Diment's practical observation and the real-world consequences of software customizability.

Second:

> But does this mean that whoever wants to use a computer must also write the software for it? Would someone purchasing an automobile for a cross-country trip first study cartography, then proceed to obtain aerial and satellite photographs of the proposed route, and finally draw a detailed map for the whole journey? Hardly. It is far easier to go to the A.A.A. and get standard maps or that organization's special trip sheets.

How could anyone have known that a scant 30 years later (2010s) that people could have a pocket-sized computer which (for the most part) would obviate the use of paper maps for navigating to unknown destinations? That entire industries supporting the production of paper maps would be dramatically scaled back because of a globally-connected infrastructure involving microprocessor manufacturing, interface design, wireless communication, and (literal) rocket science would be publicly available to nearly all comers?

Sandberg-Diment's practical answer to "does everyone need to learn programming?" is comforting, persuasive, and correct. But the impractical answer--everyone should consider learning programming-- would be to catch a glimpse of the future and the massive transformations that widely available computing would bring inside a generation.

This is the "I like tinkering with cars therefore everyone should be a mechanic" argument.

There's no justification for it. Maybe 20% of the population - at best - is even capable of that kind of programming. [1] Most people simply don't do symbolic abstraction at that level, and forcing them to try would create resentment, not literacy.

And "conlangs" are indeed different to "natlangs." There's definitely a case to be made for teaching everyone at least one extra language. But the kind of abstract thinking required for conlangs is adequately covered by basic STEM.

There might be a case for some very basic experience with programming in schools. But expecting the entire population to be able to do it at a professional level makes no more sense than expecting the entire population to have the same skills as qualified doctors, lawyers, architects, or pilots.

[1] There are fewer than 30 million developers globally, out of a population of nearly 8 billion.

> I like the bit at the end about studying the classics in Latin and Greek.

I did not see a reference to the Classics, Greek, or Latin in the article…