Slashdot Mirror
Remember the Computer Science Past Or Be Condemned To Repeat It?
theodp writes "In the movie Groundhog Day, a weatherman finds himself living the same day over and over again. It's a tale to which software-designers-of-a-certain-age can relate. Like Philip Greenspun, who wrote in 1999, 'One of the most painful things in our culture is to watch other people repeat earlier mistakes. We're not fond of Bill Gates, but it still hurts to see Microsoft struggle with problems that IBM solved in the 1960s.' Or Dave Winer, who recently observed, 'We marvel that the runtime environment of the web browser can do things that we had working 25 years ago on the Mac.' And then there's Scott Locklin, who argues in a new essay that one of the problems with modern computer technology is that programmers don't learn from the great masters. 'There is such a thing as a Beethoven or Mozart of software design,' Locklin writes. 'Modern programmers seem more familiar with Lady Gaga. It's not just a matter of taste and an appreciation for genius. It's a matter of forgetting important things.' Hey, maybe it's hard to learn from computer history when people don't acknowledge the existence of someone old enough to have lived it, as panelists reportedly did at an event held by Mark Zuckerberg's FWD.us last Friday!"
John Graham-Cumming gave a talk at OSCON 2013 titled "Turing's Curse" that speaks to this same idea. Worth a watch.
It's pretty damn obvious why this is: as an industry, we no longer shun those who should definitely be shunned.
Just look at all of the damn fedora-wearing Ruby on Rails hipster freaks we deal with these days. Whoa, you're 19, you dropped out of college, but you can throw together some HTML and some shitty Ruby and now you consider yourself an "engineer". That's bullshit, son. That's utter bullshit. These kids don't have a clue what they're doing.
In the 1970s and 1980s, when a lot of us got started in industry, a fool like that would've been filtered out long before he could even get a face-to-face interview with anyone at any software company. While there were indeed a lot of weird fuckers in industry back then, especially here in SV, they at least had some skill to offset their oddness. The Ruby youth of today have none of that. They're abnormal, yet they're also without any ability to do software development correctly.
Yeah, these Ruby youngsters should get the hell off all of our lawns. There's not even good money in fixing up the crap they've produced. They fuck up so badly and produce so much utter shit that the companies that hired them go under rather than trying to even fix it!
The moral of the story is to deal with skilled veteran software developers, or at least deal with college graduates who at least have some knowledge and potential to do things properly. And the Ruby on Rails idiots? Let's shun them as hard as we can. They have no place in our industry.
Also in line with this: I can't imagine that way patents work actually help with the problem of inventing the wheel. You almost have to reinvent the wheel to create a working solution that won't get you sued.
We marvel that the runtime environment of the web browser can do things that we had working 25 years ago on the Mac.
Did the Mac, 25 years ago, allow people to load code from a remote server and execute it locally in a sandbox and in a platform independent manner all in a matter of a couple of seconds? No. No it did not.
We should then pay homage to the Mac 25 years ago, when it basically did what Doug Englebart demonstrated 45 years ago. Nice logic you have there.
The best and brightest at Apple, MS, BeOS, Linux did learn from "the great masters" - thank you to all of them.
They faced the limits of the data on a floppy and cd.
They had to think of updates over dial up, isdn, adsl.
Their art had to look amazing and be responsive on first gen cpu/gpu's.
They had to work around the quality and quantity of consumer RAM.
They where stuck with early sound output.
You got a generation of GUI's that worked, file systems that looked after your data, over time better graphics and sound.
You got a generation of programming options that let you shape your 3d car on screen rather than write your own gui and then have to think about the basics of 3d art for every project.
They also got the internet working at home.
Domestic spying is now "Benign Information Gathering"
Indeed. Half of today's programmers have roughly zero engineering education, and want to be called software engineers. They have no idea, no idea at all, what their data structures look like in memory and why they are so damn slow. Heck "data structure" is an unfamiliar term to many.
It's not entirely young vs old, either. I'm in my 30s. I work with people in their 50s who make GOOD money as programmers, but can't describe how the systems they are responsible for actually work.
How do we fix it? If you want to be good, studying the old work of the masters like Knuth is helpful, of course. Most helpful, I think, is to become familiar with languages at different levels. Get a little bit familiar with C. Not C# or C++, but C. It will make you a better programmer in any language. Also get familiar with high level. You truly appreciate object oriented code when you do GUI programming in a good Microsoft language. Then, take a peek at Perl's objects to see how the high level objects are implemented with simple low level tricks. Perl is perfect for understanding what an object really is, under the covers. Maybe play with microcontrollers for a few hours. At that point, you'll have the breadth of knowledge that you could implement high level entities like objects in low level C. You'll have UNDERSTANDING, not just rote repetition.
* none of this is intended to imply that I'm any kind of expert. Hundreds, possibly thousands of people are better programmers than I. On the other hand, tens of thousands could learn something from the approach I described.
I'm old enough at 55 to remember the past, and yes I did love APL briefly but lamenting that the present isn't like the past is like wishing it was 1850 again so you could have slaves do all your work. Neither the web nor the modern mobile application is anything like the past, and what we use to write that code today is nothing like what I started with. Trying to relive the past is why old programmers get a reputation for being out of touch. The past is important in that I learned a lot then that still rings today but I can say that about every year since I started. Today is a new day everyday.
Lisp never 'comes back'. It merely recurses.
Actually, it was Babbage who faced such idiocy from Parliament:
On two occasions I have been asked [by members of Parliament], 'Pray, Mr. Babbage, if you put into the machine wrong figures, will the right answers come out?' I am not able rightly to apprehend the kind of confusion of ideas that could provoke such a question.
And you thought the clogged tubes thing was bad.
and he is still right. ;)
I've always felt like that quotation had another interpretation, one that's much more favorable to the MPs:
If you're an MP, you've probably had to deal with a lot of people asking for money to fund what is essentially snake oil. If you don't understand the underlying 'cutting edge' technology (both plausible and acceptable), one simple test is to ask a question that you KNOW if the answer anything other than "No" that the person is bullshitting, and you can safely ignore them... and as reported the question is phrased in such a way that it would sorely tempt any huckster to oversell their device. I think Babbage's lack of comprehension was due to his inability to understand the idea that the MP was questioning HIM, rather than the device.
There was a point made at the 30-year retrospective talk at last year's European Smalltalk conference. If you have two languages, one of which allows developers to be more efficient, then you will end up needing fewer developers for the same amount of work. Unless your entire company uses this language and never experiences mergers, then this group of developers will be outnumbered. When you begin a new project or merge two projects, management will typically decide to pick the language that more developers have experience with. If you have a team of 100 C++ programmers and another team of one SilverBulletLang programmers, then it seems obvious that you should pick C++ for your next project because you have vastly more experience within the company in using C++. The fact that the one SilverBulletLang programmer was more productive doesn't matter. In the real world, languages tend not to be silver bullets and so the productivity difference is more in the factor of two to five range, but that's still enough that the less-productive language wins.
I am TheRaven on Soylent News
That's such a cop out and it's not true. Most the managers making these decisions are technical managers who come from development backgrounds themselves.
There is a problem at a more fundamental level, even outside of determining what buzzwords to use for a product and it's prominent even in some of the higher echelons of web society. The most obvious I'm going to point out is that of HTML5 - it's a braindead spec full of utterly amateur mistakes that could've been avoided if only Ian Hickson had spent 5 seconds understanding why existing things were the way they were and why that mattered.
An obvious example is that of HTML5's semantic tags, using a study to determine a static set of tags that would be used to define semantic capabilities in a spec that was out of date before the ink had even dried was just plain stupid. The complaint that we needed more readable markup rather than div soup to make writing HTML was naive, firstly because amateurs just don't write HTML anymore, they all publish via Facebook, Wordpress and so forth, and secondly because there's a good reason markup had descended into div soup - because genericness is necessary for future-proofing. Divs don't care if they're ID is menu, or their class is comment, they're content neutral, they don't give a fuck what they are, but they'll be whatever you want them to be which means they're always fit for the future. In contrast HTML5 tried to replace divs with tags such as aside, header, footer and so forth which would be great except when you have a finite number of elements you end up with people arguing about what to do when an element doesn't fit. Do you just go back to using divs for that bit anyway or do you fudge one of the new tags in because it's kinda-loosely related which means you bastardise the semantics in the first place because we now don't really know what each semantic tag is referring to because it's been fudged in where it doesn't make a lot of sense?
The real solution was to provide a semantic definition language, the ability to apply semantics to classes and IDs externally. Does that concept sound familiar? It should because we had the exact same problem with applying designs to HTML in the past and created CSS. We allowed design to be separate from markup with external stylesheets because this had many benefits, a few obvious ones:
1) Designers could focus on CSS without getting in the way of those dealing with markup making development easier
2) We could provide external stylesheets for no longer maintained sites and have the browser apply them meaning there is a method to retrofit unmaintained sites with new features
3) Our markup is just markup, it just defines document structure, it does one thing and one thing well without being a complete mess of other concepts.
Consider that these could've been benefits for building a semantic web if HTML5 had been done properly too. The fact that Ian Hickson failed to recognise this with HTML5 highlights what the article is talking about exactly. He's completely and utterly failed to learn the lessons before him as to why inline styling was bad but on a more fundamental level demonstrates a failure to understand the importance of the concept of separation of concerns and the immense benefits that provides that was already learnt the hard way by those who came before him. His solution? Oh just make HTML5 a "living spec" - what? Specs are meant to be static for a reason, so that you can actually become compliant with them and remain compliant with them. Spec compliance once you've achieved it shouldn't ever be a moving target. That's when you know you need to release a new spec.
It's a worrying trend because it's not just him, I see it amongst the Javascript community as they grow in their ambition to make ever bigger software but insist that Javascript is all they need to do it. The horrendously ugly fudges they implement to try and fudge faux-namespaces into the language in a desperate attempt to alleviate the fact the Javascript was just neve