I saw this tweet by Julia Evans the other day.
if you've been working in computing for > 15 years -- are there fundamentals that you learned "on the job" 15 years ago that you think most people aren't learning on the job today?— 🔎Julia Evans🔍 (@b0rk) September 9, 2021
(I'm thinking about how for example nobody has ever paid me to write C code)
I've never met Julia but have been following her on Twitter and reading her blog for some time now.
The tweet got me thinking. Not so much about learning on the job but rather on how colleges prepare their students. Recent boutique majors and concentrations notwithstanding, a degree in CS seems to have largely stayed the same from the early/mid 1980s when I got my BA to today. There are variations but it was and still is largely
some systems or OS stuff
math: calc 1, calc 2, discrete, linear, maybe stats
another requirement or two which might differ based on institution
Sure, programs have changed over the years. At one point, computer graphics was hot. Now, not so much. Courses like compilers also used to be required more frequently but overall things look much the same. That said, there have been some major changes - at least from my perspective and those changes really affected how students were prepared for professional careers. Now of course this is only my perspective so who knows how accurate it is. Regardless, it should be good food for thought.
I got my BA back in the mid 80s. That was the start of the PC generation. Colleges were starting to get computer labs full of IBM-PCs and then later clones. MS-DOS ruled the day.
What did that mean?
It meant that while a few years prior, students used a time sharing system for their programming classes now they were using an IBM-PC with Turbo Pascal. I can't say for sure but looking back, it seemed that people educated in PC heavy programs had a weaker sense of all those issues that come with multiple people and multiple processes happening on a machine at the same time.
On the other hand, having a relatively simple computer - one that you had full control over may have led to an easier time in understanding the low level ins and outs. You could learn about low level issues on a PC and really understand everything. That doesn't seem to happen as much nowadays when even our personal computers and computing devices run more like those multi user computers of old than the single process no threads IBM-PC.
There was an opportunity to recapture this right before the Arduino got big. The Arduino was based on an Atmel processer with a small instruction set - a platform where a student could understand everything from wire to application. For all the good that the Arduion and related projects have led to, they've abstracted that low level away.
A few years later Java became a popular college CS language followed by Python. Both of these took memory management out of the core programming->data structures->algorithms sequence. Sure, you probably got to learn a bit about memory management and related issues in your systems classes but I have to believe you didn't really own that knowledge like you would have had you had to manage memory throughout the core courses.
Now, I'm not knocking the change. Moving to Java brought some improvements and so does moving to Python. It's just that as a language or platform giveth it also taketh away.
The interesting things here is that the changes I noticed were not directly intended but rather byproducts of changes made for other reasons.
A final unintended consequence I noticed also seems to be a byproduct of the Java revolution. Years ago after many programs had adopted Java I was talking to a friend who was a high level SE at a big NY bank. At the time finance was NY's biggest tech employer. I think Google might have recently moved to NY but wasn't yet the player they'd become - this was pre-IPO days and NY tech hadn't yet exploded. My friend was lamenting after a long day of interviews. None of his interviewees - all from top rated CS colleges could handle what he viewed as easy algorithm questions. Questions a that were considered close to trivial by his interviewees from a few years back.
After some discussion, we developed a theory. It might not be true but looking around the landscape in the months and years following that discussion, it might hold some water. With the acceptance of Java students were now, more frequently using library calls. The .sort method, the built in List, Tree, Map, and Set classes and more. This is not to say that the students hadn't learned about, say, the mergesort in their data structures or algorithms classes but rather they didn't live it. Maybe they did a quick implementation but then it was gone from their memories. Back in the day, most people in my peer group started in Pascal and frequently moved to C. Even though C had it's built in sort, we coded our own more frequently. We also didn't have easy access to all the data structure and algorithm libraries that Java had so we rolled our own and continued to use them. When kids who did that interviewed, they could more easily see how an interview problem related to an algorithm they coded and tweaked half a dozen times. On the other hand, those that saw it once and moved to a library didn't have that in depth understanding.
Now, this is not to say that it was better in the old days. When they removed memory management, they were able to add other topics. Likewise when students didn't need to implement their own version of everything for their class projects, instructors could steer students towards learning other new things.
I'm not making a judgement call on better or worse but rather that even without formal curricular changes the strengths and weaknesses of our CS graduates have changed over the years quite possibly as a result of unintended consequences of a variety of decisions.
I also don't this only happens in CS. Take a look at math - the big one would probably be bringing the calculator to the classroom. The in your face change was the idea that on the one hand, students would become weaker in arithmetic and manual calculation. On the other hand, time could be spent on what might be considered richer topics due to time saved. Lost in the shuffle were topics like linear interpolation - a useful skill taught to navigate log and trig tables. Those skills would be forever lost.
So, there's my rant and ramble inspired by Julia's post.
A day after I saw the tweet I saw this post by Julia. Unless some shiny object distracts me I think I want to look at how the unintended consequences from some of the changes I mentioned directly relates to it.