Top 10 Dead (or Dying) Computer Skills

Lucas123 writes "Computerworld reporter Mary Brandel spoke with academics and head hunters to compile this list of computer skills that are dying but may not yet have taken their last gasp. The article's message: Obsolescence is a relative — not absolute — term in the world of technology. 'In the early 1990s, it was all the rage to become a Certified NetWare Engineer, especially with Novell Inc. enjoying 90% market share for PC-based servers. "It seems like it happened overnight. Everyone had Novell, and within a two-year period, they'd all switched to NT," says David Hayes, president of HireMinds LLC in Cambridge, Mass.'"

Comments

c ? really?

[stoolpigeon]

doesn't really match up with my experience. and putting it next to powerbuilder? that's just not right.

Re:c ? really?

[WrongSizeGlass]

'C' will never die. Period. It has so many uses from PC's & 'big iron' to embedded systems.

Re:c ? really?

[fyngyrz]

No, C isn't in any way going out. C produces fast, tight code that so far, C++ and C# can't even begin to match. C++ is just C with a lot of baggage, a great deal of which you can implement in C in a completely controllable, transparent and maintainable manner. We use the most important of those regularly in C code, specifically objects and objects with methods. We obtain better performance, smaller executables, and smaller memory footprints than any company that makes similar software using C++ or Objective C's add-on paradigms. C, and the C sub-domain of C++ and so on, is no more "going away" than C++ itself is. C occupies a unique niche between the metal of assembly and the (so far) considerably less efficient higher level languages — I'm talking about results here, not code. I'm all about recognizing that a few lines of C++ are very convenient, but the cost of those lines is still too high to even think about abandoning C code for performance applications. For many, the object isn't finding the absolute easiest way to write code, but instead trying to find a balance between portability, reasonable code effort and high performance. C sits exactly in that niche. C++ is easier to write, almost as portable, but produces applications with large footprints, inherited, unfixable problems inside non-transparent objects (like Microsoft's treeview, to name one), and a considerable loss of speed as compared to a coder who has a good sense of just what the C compiler actually does (which usually means a C coder that has assembly experience, intimate knowledge of stacks and registers and heaps and so on.)

Speaking as the guy who does the hiring around here, If your resume shows C and assembler experience, you've made a great start. Even just assembler. C or C++ only, and your odds have dropped considerably. C, assembler and either a great math background or specifically signal processing, and now we're talking. C++ doesn't hurt your chances, but you won't get to use it around here. :)

Re:c ? really?

[tha_mink]

'C' will never die. Period. It has so many uses from PC's & 'big iron' to embedded systems. What is 'C'? Is that a language? Like latin?

I'm kidding, but only partially. I was a COBOL developer for lots of years, and I thought that COBOL would never die either. I would say "Too many companies are too invested ..." blah blah blah. I think that I actually even used the 'big iron' quote too when telling my friend how secure COBOL was. Um...I was wrong. I found that out with plenty of time to learn other stuff like Java and so forth but of course it's going to die. Just like C++ will die, just like Java will die, et al. If you've been in our business long enough, you should know better. Everything dies, it's just a matter of time. And if you think you're going to get 30 years out of the technologies that are new now, then you're wrong there too. That's the double edge sword that is the IT business. Keep learning, keep growing or start flipping burgers.

Re:c ? really?

[Chainsaw]

C produces fast, tight code that so far, C++ and C# can't even begin to match. C++ is just C with a lot of baggage, a great deal of which you can implement in C in a completely controllable, transparent and maintainable manner.

Wow. You must have had some really shitty software engineers. It's very likely that you can create C++ code that is as fast or faster than C. Yes, I said it. Implementing virtual inheritance and method overloading in plain C is doable, but it will be very complex. Templates? Don't even want to think about it.

C++ is easier to write, almost as portable, but produces applications with large footprints, inherited, unfixable problems inside non-transparent objects (like Microsoft's treeview, to name one), and a considerable loss of speed as compared to a coder who has a good sense of just what the C compiler actually does (which usually means a C coder that has assembly experience, intimate knowledge of stacks and registers and heaps and so on.)

I have no idea what the MS Treeview problem is, but once again - the programmers that you have worked with must have sucked balls. I'm an old C coder, with some solid x86 assembler knowledge. As you say, it's possible to get very high performing applications using C. However, why would I do that, when I can create code that is just as fast and much more readable by using C++? Yes - even for embedded development, which is my dayjob.

Re:c ? really?

[WrongSizeGlass]

And if you think you're going to get 30 years out of the technologies that are new now, then you're wrong there too. I've been coding in 'C' for 24 years, and unless OS's, drivers, embedded systems, et al, stop caring about performance I think 'C' will out last me in this industry (and probably out live me, too).

That's the double edge sword that is the IT business. Keep learning, keep growing or start flipping burgers. I've coded in over 20 languages in my career, from assembly languages to proprietary 4GL's and everything in between, on more platforms than I have fingers. My library has programming texts older than most coders today. Keep learning? Great advice. Give up on 'C'? That's another story entirely ...

Re:c ? really?

[grumbel]

### C produces fast, tight code that so far,

So can other languages, I don't think that is the main selling point of C. I think the main selling point of C is that it is by far the most "compatible" language of all. Python, Perl, Ruby and friends are all based on C, if you want to extend the languages, you do so by writing a module using their C API. If you want to simply call C functions, you can do so from many other languages as well be it Lisp, Ada or whatever. If you want to talk to the kernel you do so in C. No matter what language you use, sooner or later you come to a point where you have to fall back and either write C or interface with C code, since C is the 'real thing', while everything else is just an ugly wrapper around C code, trying to hide it, but often failing at doing so properly.

As long as a ton of stuff is based on C code it isn't going away, especially not in the OpenSource world where basically everything is based on C.

Maybe one day some Java or .net based OS will take over, but I don't see that happening for many years or decade(s?) to come.

Re:c ? really?

[jesuscyborg]

I'm kidding, but only partially. I was a COBOL developer for lots of years, and I thought that COBOL would never die either. I would say "Too many companies are too invested ..." blah blah blah. I think that I actually even used the 'big iron' quote too when telling my friend how secure COBOL was. Um...I was wrong. I found that out with plenty of time to learn other stuff like Java and so forth but of course it's going to die.

But you're forgetting the reason COBOL is dying is because the tech industry is moving away from the machines and operating systems that make use of COBOL. C will not die in the next half century because the tech industry is moving closer towards technologies that are built on C like GNU/Linux. If anything, C will become more prevalent.

In the next fifty years I imagine C's role more or less becoming that of the "mother language". 90% of the time everyone will be using higher level languages like Perl, Ruby, and Haskell on their Linux computers, all of which are programmed in C. Programmers will only need C when they need to change their lower level system tools, or to write new ones that perform very efficiently.

The only way I can see C dying is if a kernel comes along with a Linux compatible system interface that's written in a language suited better to the massively parallelized CPUs of the future. And once the kernel moves away from C, applications are bound to follow.

Re:c ? really?

[Plutonite]

You are definitely right, and it's not just because what happend over the last 24 years of your engineering history is likely to carry over..it's also the nature of the language itself. There is a reason C (and C++) are so damn popular, and the reason is that they embody most, if not all, of what can be done with a general purpose language. Things like Java and Python will stay for quite a while too, because the design there is more conforming to object-orientation while keeping most of the general-pupose flexibility, but C and the various assembly languages will never die. It would require a re-write of the entire architectural basis of computing to throw them out, and the theoretical part of computation theory does not need features that are unavailable here (yet). Anything that can be done in any language can be done (albeit less elegantly) with the aforesaid.

No one shall expel us from the Paradise that Richie has created (Apologies to David Hilbert)

Re:c ? really?

[Anonymous Coward]

Several times I've been told "nobody does assembler anymore", and yet I still keep needing to use it. I'm not writing whole applications in assembler of course. But the same reasons I keep having to do assembler are some of the same reasons that C is used. C and assembler may be less commonly used than in the past, but unlike COBOL, modern computer systems are still very heavily dependent upon them. Until we get a radically new form of computer architecture, which doesn't seem likely anytime soon, demand will remain for people who can write and maintain the guts of what happens underneath the applications. And you can't do that in Java or C# or Ruby.

Computerworld Magazine, being an IT rag, is concerned about IT, not computer science or engineering. Thus it worries about product names, not categories. So the point out skills in SNA or Novell Netware that aren't needed so much anymore, even though computer networking skills are even more popular and vital today than in the past. COBOL may be virtually dead, but dry and dusty applications for business purposes are alive and well and written by people who still wear ties.

Re:c ? really?

[644bd346996]

Sure, some things still need to be done in assembly, but they are always wrapped in a C api. And that is why C will not die until computer architectures change drastically. C is close enough to the hardware that you really know what is going on, and you can control it directly. Languages like Java, which eschew pointers and mandate that all code be in a class, can never replace C.

In fact, no language that isn't pretty much C can replace C. If it doesn't give you the control over pointers and memory allocation that you have with C, it won't work as a replacement. If it does have those thing, it is not going to replace C unless it is a backwards compatible extension like C++ or Obj-C.

Re:c ? really?

[fishbowl]

>I'm surprised that Fortran didn't make the list.

I would have been, but working in scientific research I've discovered a couple of things that are surprising:

1. People actually do calculus on the whiteboard for reasons other than taking a math class.

2. Lots of people actually use FORTRAN. Even people whose Java, C, C++, Perl, Ruby, etc. skills are such that I look up to them -- and they have solid arguments for using FORTRAN, at least for certain kinds of numerical computing.

But here's the thing: There are separate worlds. In one world, the idea of using calculus on a daily basis is simply never a consideration. You learn enough to finish college, and that's the end of it. Likewise, there's a world where numerical computing and arbitrary precision and optimized complex arithmetic are actually primary considerations and not just hypothetical things.

I never understood this until I found myself in that world. And I wouldn't have believed you if you told me that people who know other languages, choose FORTRAN even when given a choice.

But what I take from it, is that there are requirements that are met by FORTRAN which are not met by languages that offer more comfortable grammars.

People (myself included) will argue that, for instance, C can do anything that FORTRAN can do, in a much happier grammar (opinion, mine, widely shared), but the thing is... while that's strictly true, a lot of the things that seem tangential or irrelevant, turn out to be *crucial*, where seriously optimized math support is the core of the application. FORTRAN makes guarantees on the kinds of things that are implementation dependent in C.

Anyway, there's no shortage of FORTRAN programmers. It's quite easy for a skilled programmer to learn FORTRAN, once you get past the 'WTF' factor and can accept that it's relevant in todays world, at least when your problem space is a good fit for the language.

COBOL or PL/1 and the like, make another story entirely. My experience has been that the role of COBOL has been replaced by the combination of transitions to modern RDBMS, decentralized business processes as a side-effect of the whole ubiquitous "PC" adoption, and the adoption of, for example, Enterprise Java. That covers one end of the spectrum, and the other end (the big corporate end) is covered by the evolution of vertical systems providers (e.g., Peoplesoft, SAP, SAIC).

Back on topic: If there's a university CS program that gives degrees without courses in Operating System and Compiler design taught in C, I'd love to hear about it. No way are C programmers in decreasing supply. If nothing else, the million or so open source projects have created a whole generation of self-taught folks who know C.

Re:c ? really?

[atomicstrawberry]

When you say 'based on C', do you mean that the compiler / interpreter is written in C, or that the language itself is derived from C? Because technically Ruby is based off Perl and Smalltalk, not C. The Perl side of things can be traced back to C, but Smalltalk's origins are in Lisp.

However they are all implemented in C, as is PHP. In fact, I'm reasonably confident you'll find all of the web languages that the article declares are taking over are implemented using C. As is Apache, which is the backbone of the majority of internet servers. In fact, pretty much everything that provides important infrastructure is written in C.

There may be demand right now for programmers that know the latest fad high-level language, but the demand for competent C programmers has hardly disappeared. The only reason that C would die is if another fast, portable, general-purpose language like it came along that offered significant benefits over C. I can't personally see that happening any time soon.

Re:c ? really?

[Khashishi]

Any programmer worth his salt can pick up a new language in a couple hours. Hell, most languages today are just ALGOL with some syntactical refinements, and you know one, you know them all. I'm not worried if Java or C or Matlab dies out. What separates programmers is ability, not language experience.

Re:c ? really?

[Kjella]

Any programmer worth his salt can pick up a new language in a couple hours.

If you mean whether it's curly braces or brackets or none at all and the syntax of basic control flow, then yes. If you mean being familiar with the standard library, the development tools and all the specific bits (Java generics, C++ templates, take your pick)? No.

Re:c ? really?

[shutdown -p now]

Any programmer worth his salt can pick up a new language in a couple hours. Hell, most languages today are just ALGOL with some syntactical refinements, and you know one, you know them all.

Surely you meant to write, "all ALGOL-family languages"? Which includes most mainstream ones: C/C++, Java, C#, Python, Perl, BASIC etc. But you can't easily jump from C++ to Lisp, Erlang, Prolog or FORTH (just to name a few) easily, because they are different. Then there are people having troubles moving from class-based OOP (Java) to prototype-based (JavaScript). Etc... there is a lot out there, and it does differ.

Re:c ? really?

[Sir Toby]

I think you will only need to wait till 2100

Or 2050, when all of the following logic begins to fall to pieces:

if ( x > 50 )
      return 1900 + x;
else
      return 2000 + x;

Some of the list looks good

[Sycraft-fu]

But C? Really? I guess that the fact that nearly every game, every OS, almost every high performance computation tool and so on are written in it (or C++ which I keep under the same heading) doesn't count. While it certainly isn't the be-all, end-all, it is still widely used. Even games that make extensive use of scripting languages, such as Civilization 4, are still C/C++ for the core functions.

Until there's enough spare processor cycles that it really doesn't matter how much CPU time you use, or a managed language gets as good at optimizing as a good C compiler/programmer combo (unlikely) I don't think C is going anywhere.

Re:Some of the list looks good

[LWATCDR]

C++ is still alive and well.
I think they are wrong since C is still used on a lot of embedded systems where C++ is too heavy.
BTW a good number of HPC tools and applications are still written in FORTRAN.

ColdFusion Dead?

[AKAImBatman]

I can only hope. Terrible, terrible language. Of course, these days it's actually a template engine for a J2EE server. So it's not nearly as bad as it once was. Unfortunately, most of the ColdFusion projects are massive, sprawling directories from the CF4/CF5 days. You're not likely to see a nicely package JAR here. :-/

Also, what's with "PC Network Administrators"? TFA must be referring to a rather specialized form of administrator, because last I checked we still needed someone to keep the desktops configured, the networks running, the file severs sharing, the login servers logging people in, and the IIS servers serving.

Wait, there IT people who specialize this much?

[Rakishi]

I mean, this is IT where things change quickly and at times unexpectedly. If you don't have at least a number of diverse skills then I can't say I feel sorry for you when your job gets axed. I may not be a guru in any one language but at least I won't be unemployed when that language dies out.

C? You must be kidding

[Ckwop]

As the Web takes over, C languages are also becoming less relevant, according to Padveen. "C++ and C Sharp are still alive and kicking, but try to find a basic C-only programmer today, and you'll likely find a guy that's unemployed and/or training for a new skill," he says.

What the web can now allocate memory and talk to my hardware? Even if you're not a kernel programmer, the web has sucked and still sucks for application development. It will continue to suck for years, due to Internet Explorer. It's misleading to claim AJAX will solve all these problems because it won't. In fact, it might even cause a few problems of its own. For example, do you really think all that AJAX is secure? In short, I think the web is taking over what naturally comes to that medium. It is wrong to say its displaced C.

Does this guy forget that all of the GNU/Linux Kernel base system is written in C? You know, the operating system that powers most web-servers? I'll tell you one thing, C will still be here in twenty years time when Ruby on Rails is talked about much in the same was Blitz Basic is today. C is here to stay; it's immortal.

Simon

10 more dying computer skills

[Anonymous Coward]

1. secure software coding
  2. data management theory
  3. data modeling
  4. usability
  5. interface design
  6. use of testing, version control, refactoring, and other best practices
  7. space or time efficient algorithms
  8. general communications skills
  9. basic business concepts like ROI
10. business ethics

They said something else.

[khasim]

"C++ and C Sharp are still alive and kicking, but try to find a basic C-only programmer today, and you'll likely find a guy that's unemployed and/or training for a new skill," he says.

Now I know some people who've learned on C#, but I'm sure that will change in the near future.

Anyone who originally learned C, and is still writing code, has probably picked up a few other languages over the years.

Re:They said something else.

[rs79]

"Anyone who originally learned C, and is still writing code, has probably picked up a few other languages over the years.

I learend Assembly first and have programmed nearly every cpu except a vax. I learned C second (in 1976) and still use it every day.

I also know: apl, forth, snobol, fortran, rpg, cobol, lisp, smalltalk, pascal, algol and probably more I can't remember. I've never done much in them except for a tiny bit of forth. I know postscript pretty well and used it quite a bit. But I do C, day in day out and do not think very much of C++ or (worse) C#. The oddball languages have their place but for most it's a pretty narrow niche. I'd rather invert a matrix in Fortran than C though, I'll admit.

Andrea Frankel said it best on usenet in the 80s: "If you need to hire a programmer ones with assembly are better than ones without".

It worries me that people today don't actually know how computers work inside any more.

Re:They said something else.

[Opportunist]

To quote my Guru: "When you learned C, and you mastered it, you have learned every (procedural) language there is, for it is easier to take a step down rather than up."

It's pretty much true. Look at the other languages you "should" learn today. Perl, PHP, Python, C#, Java... When you know your C well, learning them is fairly easy.

Re:They said something else.

[fishbowl]

How do you get out of university without taking an architecture course that gives some assembly language, at least for a hypothetical machine?

If you claim to have a BS in CS at the interview table, but didn't suffer through, e.g., a computer organization course (like Hennessy and Patterson style, which is common these days), didn't have a course where you developed an operating system, didn't design a language starting from BNF and build a compiler for it, didn't take 2 years of a lab science, didn't at least come close to a math minor, didn't have at least 4 courses at various levels of discrete math, automata, algorithm analyis, and didn't have a course that, as a final project, you deliver a significant user app in a high level language (on the order of an original multiplayer game, let's say)... I'd say your school has some explaining to do.

Seriously, what school can you go to and somehow avoid having a significant background in several languages and paradigms, including but certainly not limited to asm, C, and either C++ or Java, if not both? I don't expect everyone to take an elective where they compare Lisp, Scheme, Ruby, Haskell, and Icon. And I realize that there's often a variety of senior-year choices, and some people don't take Databases choosing HighPerf/Parallel/Distributed computing, say, or 3D Graphics. But there are some basic things that you'd better have done, or else, despite having the piece of paper hanging on your office wall, you're not done with school!

Web Design

[happyfrogcow]

Judging by their web page, all design jobs are dead too. We should all just write web pages to serve ads, because C is dead.

This article is trash, even if it does have some technologies that are irrelevant. It has very little value to the reader. I'd rather read a 10 top list for reasons Paris Hilton should be locked up for life.

Re:They are nuts on the C front.

[Dionysus]

a C programmer can move to C++ without a problem but the reverse is not true I find the opposite to be true. A C++ programmer is able to move to C without much problems, but the oppose it just not true.

Re:You have to admit

[ivanmarsh]

Yep... after all everyone knows that C# is the best language with which to progarm an embeded micro-controller.

Technology reporting is certainly dying.

It's not C. It's the C only programmer.

[khasim]

They phrased it very badly. C isn't going anywhere. But if all you know is C, then you are very rare.

Most programmers who know C also know at least one other language.

In any event, putting that on the list was just stupid.

Re:dovetail

[MightyMartian]

I don't think you can justify C and Cobol. There are millions upon millions of lines of code in these two languages, and despite all the sexy new ones that have come along, these two still reign supreme; C is incredibly prevalent on dedicated systems and within a lot of operating systems, and mainframe Cobol code can still be found throughout the business world (though often cleverly disguised these days). I doubt a skilled Cobol programmer will be at risk of starving any time in the near future.

PC network admins?

[Volante3192]

With the accelerating move to consolidate Windows servers, some see substantially less demand for PC network administrators.

Apparently this guy's never dealt with users. If there's a way to screw up a system, even a dumb terminal, they WILL find a way.

Yet more netware bashing...

[damn_registrars]

I guess this shouldn't surprise me. Even though Netware was pulling "five nines" (of reliability, for those not familiar with the term) long before anyone considered running any flavor of windows on a server, we see another article bashing Netware.

Sure, its sales have declined drastically, but I wouldn't say that its relevance has. I'd be willing to bet that if we were to actually survey what file servers are still running out there, we'll see a much larger representation of NetWare. Just because people aren't buying the latest version doesn't necessarily mean that they aren't using the old ones.

For two years, I managed the computer network of a daily newspaper - including through the election debacle of 2000 and the 9/11 events. We ran that network primarily off of four netware 4.11 (later netware 5.0) servers. One of those servers had been running for over 400 days continuously when I left, and it served files and print jobs. That kind of reliability is hard to match.

Then Pay the Market Wage

[Anonymous Coward]

If you pay the market (equilibrium) wage, then you will find plenty of workers. However, most companies, just like your company, refuse to pay the market salary. They then cry, "There is a shortage of workers!"

Re:Raising the bar

[AuMatar]

You don't master pointers til you learn assembly. Until then, you just don't truely understand addressing and memory use.

10 undead computer skills

[QuantumG]

Everyone says they are dead, but they just won't go away!

1. functional programming
2. formal methods
3. prolog
4. LISP
5. Scheme
6. Smalltalk
7. Pascal
8. Tcl/Tk
9. LALR parsing
10. pre-bash shell scripting.

and that's the real message here.. nothing is thrown away in computer science.. we're just too damn young a field to honestly say we've hit a dead end on any particular technology. Anything you can name, people have done work on it in the last 10 years.

I'm more concerned with dead USER skills

[sootman]

1) knowing what extensions are
- Both the fact that that they exist in the first place AND what the different ones mean--"ooh, should I click on hotsex.jpg.doc.exe.scr.pif?"

2) looking at the URL in the status bar before clicking on a link
- Apple: I love you, but you SUCK for having the status bar off by default in Safari.

3) knowing where downloaded files go
- Every phone-based support call I've ever made:
a) Painfully (see #4) navigate to a URL.
b) Painfully (see #5) instruct user to download a file.
c) Spend 5 minutes telling them where that file is on their computer

4) the difference between \ and /
- these people saw a backslash ONCE in their lives while using DOS about twenty years ago, and now every time I tell them an address, it's "Is that forward slash or backslash?" (Despite the fact that I've told them a million times that they'll pretty much NEVER see a \ in a URL.) This is usually followed by the question "Which one is slash?" God damn you, Paul Allen.

5) the difference between click, right-click, and double-click
"OK, right click on My Computer... no, close that window. Now, see the mouse? Press the RIGHT BUTTON..."

6) the concept of paths, root directories, etc.
- I why do I have to explain fifty times a day how to get from example.com/foo to example.com?

Admins can get whatever skills they want--they picked the career, thy can accept the fact that things change. The backends are usually handled by people with some know-how. It's the end-users that cause all the problems. It'd be like driving in a world where people didn't know how to use turn signals, didn't check their blind spots, didn't know they shouldn't talk on the phone while making complicated maneuvers--oh, wait, bad example.

I highly disagree with number 9!

[Opportunist]

A good network admin is sought after. And he will never be out of a job.

Notice the "good" in the above statement, please!

Unfortunately, network admins have already suffered for years from what we (programmers) are facing now: Clueless wannabes flooding the market. Sounds harsh, is harsh, but it's sadly true. Everyone who can spell TCP/IP and doesn't think it's the Chinese secret service calls himself a net admin. And since human resources usually can't tell a network cable from a phone cable, they hire the ones with the cutest looking tie. Or the one with the most unrelated certificates.

Quite frankly, I have met so many people who claim to be net admins who know even LESS about networks than me. And I can barely cable my home net, and I can't solve the retransmission issues with my game machine that clog it. I do expect a lot from a net admin, granted, but for crying out loud, it's their JOB to know more about networks than I do, or I could do it myself!

What you get today as a "network administrator" is some guy who can somehow, with a bit of luck, good fortune, a graphical interface and a step-by-step guide from the 'net, get the DHCP server on a Win2003 Server up and running. Don't bother trying to get a static IP or even a working DNS server from him. Not to mention that he'll look blankly at you when you ask him about splitting the 'net into smaller chunks. Anything in a netmask other than 00 or 0xFF (sorry: 0 and 255) is alien to him.

That's not what I call a network administrator. That's what I call a clickmonkey.

True network administrators who got more than an evening school degree are still rare. And they will have a job, with companies that know what to look for in a net admin.

But the plague spreads. Recently we hired a "programmer" who doesn't know the difference between heap and stack. Or why inserting an inline assembler line of INT3 could do some good for his debugging problem.

And we wonder about buffer overflow issues and other security problems in code? I stopped wondering.

Re:dovetail

[Watts Martin]

I think you (and many others) are somewhat missing the point of the article, although the somewhat histrionic headline encourages a "miss the forest for the trees" reading.

I don't think anyone is expecting C or even COBOL to vanish with the speed of PowerBuilder or NetWare; the issue is whether those are actually "growth markets" any more. The article is asserting they're not, and particularly in COBOL's case I'm pretty sure that's correct. COBOL will probably live on for quite some time, but you don't hear much about people deploying new COBOL projects -- you hear about them supporting existing ones that haven't been replaced.

As for "but the OSes are written in C!" as a battle cry: well, yes, they are. But 25 years ago, they sure weren't: C was just too damn big and slow to write an operating system in. What's happened since then? Computers have gotten orders of magnitude faster, RAM and disk space have gotten orders of magnitude bigger, and of course compiler technology has also just gotten better. Couple that with the fact that operating systems and the computers they run on are just a lot more complicated -- having a higher-level language to deal with that, even at the system level, is a real advantage. There's nothing that prevents you from writing an operating system in assembly language now, but under most circumstances you probably wouldn't want to.

The thing is, unless you want to assert that computers twenty years from now will not be much faster and have much more storage and be much more complicated, you can't assert that moving to a higher-level language than C will never be either practical or beneficial even at a system level. I don't expect C to go away or even be relegated to "has-been" status, but I suspect in the long term it isn't a growth skill. It's going to move more deeply into embedded systems and other arenas where small is not merely beautiful but necessary.

The comparison with COBOL may be overstated, but it may not be completely inapt: the fact that there are still COBOL jobs out there and they may actually be fairly high-paying ones doesn't mean that going to school, in 2007, in preparation for a career as a COBOL developer is a bright idea. The same isn't as true for C, but I'm not convinced that's going to stay true for that much longer, let alone indefinitely.

COBOL is not a skill

[glwtta]

Neither are C, ColdFusion, or NetWare certification - programming and software design are skills, as is network administration; what they list are called tools.

Re:I agree totally.... BUT

[adamjaskie]

I don't understand this whole "computers are faster; why bother making things run fast?" thing. Why can't we keep writing efficient code, run it on the faster modern machines, and have things actually GO FASTER? It seems that as computers get faster, application programmers get lazier, and everything runs at the same pace. What used to take 20 cycles now takes 4000 cycles, but those 4000 cycles happen in the same time as the 20 cycles. Is that an improvement? Not in my book.

Re:I agree totally.... BUT

[swilver]

Actually, I think it's not as bad as you think -- you assume that a 10 times faster CPU actually performs 10 times faster in every way, unfortunately it usually means that the CPU wastes 10 times more cycles waiting on stuff. Yes, there are lots of things that used to take 20 cycles that now take 4000 cycles, but mostly those things are not time-critical anyway (like user interaction, anything that has to access the network or harddisk -- webservers spring to mind). For example, does it really matter if my GUI code runs in 20 cycles vs 4000 cycles? No, it won't make any difference at all -- the bottle neck here is how fast a user can click (and how fast the graphic card can render all those "special" effects users seem to love these days). Optimizing that code to run in 20 cycles won't give any performance increase at all -- optimizing it would not only be a waste of time, but would be like fixing something that already works -- the best case you can hope for is to get a faster piece of code with as much bugs as the original.. usually however it will introduce new bugs.

Now let's take this a bit further -- how much of a performance hit do you take when you access memory that is not in the CPU's cache (or 2nd level cache)? The CPU will have to wait for the memory to be available... optimizing code that frequently accesses memory outside the cache would be useless (and would just mean the CPU has to wait a bit longer). Let's take quicksort, the algorithm isn't particularly hard but accesses memory a lot. Would it matter if one iteration takes 20 cycles or 40 cycles on a modern CPU (let's assume that's the difference between C and Java)? It will make little difference, the CPU has to wait for data anyway. In the end, even in such a low level algorithm, it will make little difference whether we used a very efficient piece of code, or a slightly less efficient one -- the bottleneck is the memory. In other words, as long as the algorithm you use is the same, both pieces of code should be about as efficient.

The only time optimizing is still worth it is when you are doing stuff in tight loops that isn't randomly accessing memory for all kinds of reasons (and which of course is used to do a lot of bulk processing, like video encoding) -- it's hard to even think of a good example, but I suppose it might be worth using more efficient code in signal processing, compression/decompression and rendering applications. Even in those cases however a lot of stuff is handled in optimized libraries for higher level languages.. I mean, it won't make any difference if I use Bash (horribly inefficient!) to call my favourite Unzip program to unzip a multi-megabyte file, or whether I wrote a C program to do the same. It would still take as long.

BAH - 10 already-dead skills...

[bscott]

0. Tweaking IRQs on PC clones to let soundcards work with any other card
1. Knowing how to drop certain types of home comupter to re-seat the chips
2. Inserting 64k RAM chips with your bare hands to expand memory
3. Cutting a notch in 5-1/4" floppies to use the other side
4. Adjusting graphics by hand to NTSC-legal colors for decent video output
5. Editing config.sys to push drivers into HIMEM in order to free up memory
6. Crimping your own RJ45 connectors to save money
7. PEEK and POKE locations to do cool stuff on the Commodore 64
8. Manually configuring a SLIP connection to connect to the Internet (in pre-Winsock days)
9. Removing adjectives and punctuation from code comments to fit into 1k of RAM