I Don't Agree That There Really Is a Tech Worker Shortage

I'm a consultant, I take the work that the clients offer.

Most commonly I'm asked to write the Macintosh equivalent of some existing Windows program. While I know lots of stuff about Mac and Windows, it's quite common that I don't have a clue about anything else that that program requires, other than the core OS and standard library calls.

The way it commonly works is that I'll apply for a consulting gig. If I'm asked to interview, I'll buy the required book after I have agreed to interview, but before the interview takes place. That means I only have a few days to read the book.

You'd think that would not work but actually it works just fine.

What does not work at all is when someone wants to know how many years of experience I already have with a given technology. "None at all, but I can buy an O'Reilly book for just thirty bucks."

"a tech stack they are already familiar with".

Like 16-bit assembly code?

PC-NFS?

How to diagnose a serial cable problem with a breakout box and an oscilloscope?

How to calibrate a CCD image sensor?

So how are today's iOS coders going to feed their hungry children when no one remembers what an iPhone even was anymore?

I quite commonly point out that I can debug anything. I'm also quite good at performance optimization as well as reverse engineering.

The API, programming language, platform or what have you - the tech stack - are all orthogonal to my being one of the very best in the business at debugging.

It's not like no one needs their code debugged, but they don't seem to recognize that until it's way too late.

What they ask for is "How many years of Python experience do you have?" If I say "Less than one" - which is what I actually have - I don't get an offer. Were I able to claim that I had five years experience, I'd have a job.

What I'd like to see, is an interview where someone handed me a buggy Python program, then asked me to fix it.

While that does happen, it is quite rare.

"(Github, App store..."

Quite commonly I am asked for my Github link, but then regarded as unqualified because I don't have one.

That's because you can download my source from my own website.

I have some pretty good reason to believe that many of those who want to know my Github, don't know what Github actually is. That is, not only do they not inspect anyone's Github source, if pressed, they could not explain the difference between a github and a hubcap.

Almost universally, one is required to have at least on App in the App Store or Google Play.

There's no specific requirement as to what the app actually does, whether it works well, whether it's source code makes sense to anyone, whether it gets good reviews - just that one just that that special magic ticket into a mobile development job.

Again, I have good reason to believe that few ever actually so much as look at the app store or google play pages, let alone download, install and run the app.

Have you ever heard someone say "I'd like to work with you so we can find a way that you can drive a brand new car right off the lot. How about I give you a thousand dollar trade-in on your totally thrashed beater?"

I actually had a Toyota saleswoman say just that to me. The reason I was shopping for a new car, is that my thousand dollar trade-in was so unreliable that I was to be fired if I didn't buy a new car Real Soon Now.

Now consider that many recruiters are really good at just that kind of sales.

It's not that they try to sell anything to the candidates. They sell their, uh, "service" to the hiring managers, by somehow convincing them that its worth paying them tens of thousand of dollars in commissions for placing just one coder.

My observation is that quite a lot of folks know all about computer science, they might well know how operating systems work, but are clueless as to what happens when they combine the two.

Consider that with certain exceptions, QuickSort is widely regarded as the very best way to sort.

What happens when you sort a memory buffer on a Mac OS X box? Well that depends on a whole lot of things. How big is the buffer? If it's really large, you will cause paging no matter what you do to avoid it.

However there are all kinds of things you can do to keep that paging to a minimum. Not only does that make your code faster, it saves on electrical power and reduces the heat produced by the box.

That kind of thing is the main focus of my work these days - how to make existing code work better on modern computers.

Have a look through the source of, say, glibc, while keeping in mind what code and data memory caches do, paging, context switching with other programs and so on.

I expect that glibc would work really well on a box with no cache at all, and enough memory that it never has to page, and with only one executable running at any one time.

But that no longer applies.

Unfortunately I've never even heard of a school where they teach that kind of stuff.

Many of my colleagues regard me as delusional for asserting that one can save energy by refactoring one's source code, yet they don't seem to clue in to the fact that a computer needs to have a battery or a power cord.

"Don't tell the user to click the OK button. Tell the user to click the OK button until it turns black." -- David Courtney Johnson.

All too commonly, the work I'm actually paid to do is to make a button turn black when someone clicks on it.

That might actually be interesting, were it not for the fact that the only way to make that button turn black, is to allocate ten megabytes of objects, then run a few hundred thousand lines of code, at which point a hardware GPU writes zeros all over a memory buffer.