alt Hacker News> I have tried to convince people that ASM is reasonable as a first stage teaching language.
Unless you're teaching people preparing for engineering hardware perhaps, I think ASM is absolutely the wrong language for this. The first reason is that programming is about problem solving, not fiddling with the details of some particular architecture, and ASM is pretty bad at clearly expressing solutions in the language of the problem domain. Instead of programming in the language of the domain, you're busy flipping bits which are an implementation detail. It is really a language for interfacing with and configuring hardware.
The more insidious result is that teaching ASM will make an idol out of hardware by reinforcing the notion that computer science or programming are about computing devices. It is not. The computing device is totally auxiliary wrt subject matter. It is utterly indispensable practically, yes, but it is not what programming is concerned with per se. It is good for an astronomer to be able to operate his telescope well, but he isn't studying telescopes. Telescope engineers do that.
Replies
Ooh, very much disagree with a lot of these assertions. The problem I always encounter when trying to teach programming is that students completely lack an understanding of how to imagine and model the state of the computational system in their heads. This leads to writing code that looks kinda like it should do what the student wants, but betrays the fact that they really don't understand what the code actually means.
In order to successfully program a solution to a problem, it is necessary to understand the system you are working with. A machine-level programming language cuts through the squishiness of that and presents a discrete and concrete system whose state can be fully explained and understood without difficulty. The part where it's all implementation details is the benefit here.
> The first reason is that programming is about problem solving, not fiddling with the details of some particular architecture, and ASM is pretty bad at clearly expressing solutions in the language of the problem domain. Instead of programming in the language of the domain, you're busy flipping bits which are an implementation detail.
"How do I use bits to represent concepts in the problem domain?" is the fundamental, original problem of computer science.
And to teach this, you use much simpler problems.
> ... reinforcing the notion that computer science or programming are about computing devices. It is not.
It is, however, about concepts like binary place-value arithmetic, and using numbers (addresses) as a means of indirection, and about using indirection to structure data, and about being able to represent the instructions themselves as data (such that they can be stored somewhere with the same techniques, even if we don't assume a Von Neumann machine), and (putting those two ideas together) about using a number as a way to track a position in the program, and manipulating that number to alter the flow of the program.
In second year university I learned computer organization more or less in parallel with assembly. And eventually we got to the point of seeing - at least in principle - how a basic CPU could be designed, with its basic components - an ALU, instruction decoder, bus etc.
Similarly:
> It is good for an astronomer to be able to operate his telescope well, but he isn't studying telescopes.
The astronomer is, however, studying light. And should therefore have a basic mental model of what a lens is, how lenses relate to light, how they work, and why telescopes need them.