The Case For the Blue Collar Coder

theodp writes "U.S. tech talent shortage discussions tend to focus on getting more young people to go to college to become CS grads. Nothing wrong with that, writes Anil Dash, but let's not forget about education which teaches mid-level programming as a skilled trade, suitable for apprenticeship and advancement in a way that parallels traditional trade skills like HVAC or welding. Dash encourages less of a focus on 'the next Zuckerberg' in favor of encouraging solid middle-class tech jobs that are primarily focused on creating and maintaining tech infrastructure in non-tech companies. Dash also suggests 'changing the conversation about recruiting technologists from the existing narrow priesthood of highly-skilled experts constantly chasing new technologies to productive workers getting the most out of widely-deployed platforms and frameworks.'"

Germany - 1960's

[MadMaverick9]

They've had that in Germany since the 1960's.

http://de.wikipedia.org/wiki/Mathematisch-Technischer_Assistent [wikipedia.org]

Re:Germany - 1960's

[Anonymous Coward]

A highly misleading post, not least since the link is in German.
Yes, MTAs have been around for quite a while, but they require either a "Realschul" Diploma or a full-blown Abitur (says so in the linked article), which more often than not includes material that's usually taught in freshman years at colleges in the States.
So, it's back to the main point, namely, that you need to go to college for at least a year in the States. Then you might as well finish your CS degree.

Floating point in financial transactions

[sjbe]

Huh? Floats is the standard representation of numbers almost everywhere.

Not in financial transactions. While there are ways to do financial transactions with floating point numbers, they have an alarming tendency to introduce rounding errors. When you are dealing in money, rounding errors are an extremely bad thing because then the books don't balance anymore. One common way to deal with the problem of rounding floats is to treat the stuff to left of the decimal as an integer and the stuff to the right as another integer since there are no rounding issues with integers. While not as fast as floats, the extra accuracy is worth it in this instance. There are other ways to solve this problem but you'll find conventional floating point is used with great caution in the financial world.

There's a reason floats are implemented in hardware.

Which has nothing to do with why floating point numbers are often not used for financial transactions.

Lying abstractions

[betterunixthanunix]

Floats are an example of what I have started to refer to as "lying abstractions." When programmers deal with floats, they are being told by their programming language that they are dealing with a representation of real numbers, and the only way to avoid the situation GP described is to expend mental effort remembering that a float only has some fixed number of bits. When you have a tight deadline, you are going to spend your mental energy on other things (hopefully program logic) and it is inevitable that somewhere, you will forget that floats are not arbitrary precision.

The better answer is to use languages with arbitrary precision representations of rational numbers. You will not have to deal with integer overflows, you will not have to deal with losing precision, and you can spend more of your time developing the correct program logic. Yes, it will mean a bit more in resources -- which can be fixed later if it becomes a problem.

There's a reason floats are implemented in hardware.

Yeah, and it is the same reason we continue to use C and C++ to write high level programs: history. If we were starting from scratch today, using the latest technologies and research, it is almost certain that we would do things differently.