The USS Quad Damage

Sunny,

I agree with the first part of your argument. I don’t think Jeff Atwood got the point.

Michael Braude went on to refute some of Jeff Atwood’s points.

I think he addresses your point that some smart people are doing web-development. However, he doesn’t seem to want to have to deal with the dross that goes with them.

However, I do have some quibble with some of your quibbles.

• The first is purely a matter of definition, but I would argue that user interface is a key part of Software Engineering. It may not be a key part of
  “programming”.
• I suspect the original mention of compilers was related to writing them, which is traditionally considered a complex and deep part of software development. The fact that GWT exists doesn’t invalidate that point.
• You are right that concurrency is critical to the web, but it has been grossly simplified for mass consumption. A vast majority of projects use exactly one concurrency model – a separate process per HTTP request (okay, pooled, but the developer doesn’t need to understand that), with the database back-end providing transactional locking. It is a lovely model – versatile, simple and highly abstracted – but it means most developers don’t have to even consider other models.
• I hadn’t heard of Quake Live before. Wow! Cool. Too bad it doesn’t run cross-platform (won’t run in my browser). In any case, I am not sure you invalidate his point that the pool of web-developers doesn’t include many people who can do 3D programming. (Not that I ever needed it as a developer of applications, servers or embedded systems!)
• I would argue that at least some of the concepts in his list aren’t C++ specific (virtual methods, pointers, pass-by-reference vs. pass-by-value) and others aren’t C++ at all (garbage collection, finalizers).
• .NET has “net” in its name because it was a marketing fashion to reference the Internet, even when it wasn’t relevant, so I don’t think that proves anything.
• “any non-trivial project needs a project” – sure, but the proportion of trivial web-based projects is (arguably) higher than the proportion of trivial application-development projects. Also, arguing that Agile is suited for web-development isn’t that same as arguing Agile is commonly used for web-development projects.

Web-development may well be hard, but once of the time-consuming issues which always seems to bog us down is the difficulty of making the application run on all the different platforms out there. It is a man-made maze. It isn’t as fun to explore as the challenges that are raised by the real problems the customers are having.

Hi Julian,

Jumping right in, yes, the first is a matter of definition. I see software engineering the same way I see electronic engineering or mechanical engineering, which involves very little UI stuff. When people start dying from computer “glitches” we’ll all start seeing it that way, but I’m willing to believe that software engineering is a different beast.

Compilers, as far as I know, have all been written already, and there are well known process and algorithms to write them. You’re treading old ground if you write a compiler. I can understand that it’s a rite-of-passage of sorts for computer-ey dudes, but once it’s done, you don’t need to do it again, unless you’re writing a programming language (and there’s enough of them around!)

I agree with your concurrency point. In fact, I was going to go into more depth with how, yeah, concurrency issues are either washed over with grossly pessimistic systems, or with grossly optimistic systems, both of which leave whatever app you’re building being either slow or buggy (or both). The real concurrency issues plaguing the web haven’t been solved (unlike compilers) and someone needs to come up with these. I’ll talk more about doing stuff badly below.

Michael’s follow-up article kind of leaves us at odd ends of the chain. Mike sees web development as primarily front-end work, whereas I see web development as both front-end and back-end work, which hooks into powerful desktop, or mobile, apps. The concurrency story is completely different from that perspective.

The statement was “They don’t understand… 3D” which, clearly, some do. I understand that it stretches the definition of “web”, but there’s certainly a bunch of 3D stuff out there (go to the three website to look at phones in 3D, some car manufacturers let you look at their cars in 3D). In any case, 3D is similar to compilers: It’s (mostly) done, rite-of-passage stuff.

There are languages other than C/C++ which use pointers, and pass-by-reference and pass-by-value, and specifically mention “virtual functions”? Man bad times. As for garbage collection, there are a lot of Java programmers that don’t understand that… I’ll talk more about this below.

While I understand that it’s just a marketing term, they are trying to set up a tenuous link with the word “.NET”, like “hey maybe you’d like to use it for net programming”. The intention for the language is clear.

I take your final point regarding process. However, all of this is kind of a waste of time arguing, because what I’m trying to say is almost tangential:

In a web programming context, doing stuff well is fun and challenging. Doing stuff badly is not. If you just say “all of the bad web programmers suck, and I don’t want to be like them”, you can remove the word “web” from that sentence and the statement remains as true.

I’m over a lot of the puritanism with coding. If some guy is making a crust doing something badly, good for him or her. Basically, they’re conning themselves and their customers, but sometimes this is necessary to bootstrap the process. We shouldn’t stop things from becoming more abstract and easier for us just because it lowers the bar of entry for others. I understand that most normal folk can’t tell the difference between decent software engineers and dodgy hackers, but this will change over time — it took a long time for doctors and other engineers as well.

I was nodding in agreement with you as I read your reply.

• It’s fine for us to have different preferred definitions of SE, as long as we know which one we are using. I wonder though, what you call the field of software development that includes user experience design. (Also, people have already started dying from computer glitches. We passed that milestone decades ago.)

• “there are well known process and algorithms to write them.” Sure, and I think Michael’s point was “And the web-guys don’t know them.” I don’t think his argument holds any water. So what? Neither did Einstein. If, as you suggest, compilers are merely a rite-of-passage (which is a little too dismissive of a significant portion of my career!) then Michael is complaining that the typical web-developer hasn’t past that milestone.

• The different definitions of web-development that you highlight (front-end versus back-end) is, I think, key. Google Search is a key part of the web, and yet the amount of smarts that is spent on the front-end pales in comparison to the back-end. Is that web-development? How can it not be?

• I agree with your 3D assessment. It is a big “so what?”

• “languages other than C/C++ which use pointers, and pass-by-reference and pass-by-value” Sure. Pascal, BASIC and Ada all spring immediately to mind. (Yes, yes, Ada uses Copy-In/Copy-Out semantics. Close enough.)

• “specifically mention ‘virtual functions’?” Hmmm… That’s a bit trickier, but doesn’t Java have an equivalent concept, if not the same term?

• Yeah, your rebuttal about .NET is fair. I take your point.

• “you can remove the word ‘web’ from that sentence and the statement remains as true.” I LOLed. You are right; Sturgeon’s Revelation is universal, not just the web.

However, in an attempt to see Michael’s point of view: the communities associated with some technology appear to have more or less morons per square metre, in a way that is independent of the actual technology. There is nothing wrong with the technology that YouTube uses to stream video, and yet the YouTube commenters are notoriously lacking in thought. Visual Basic isn’t a bad language, per se, and yet its community is parodied by others. The Java community has slower migrated from (in my experience, often obnoxious, but definitely capable) early adopters to a huge community of code-churners, leaving the early adopters straining to move away from it.

Michael seems to be attacking the current web-development community, which has turned into code-churners. Jeff Atwood who left VB fairly late, seems to be a champion of the code-churners who want to be good at their job.

• “it took a long time for doctors and other engineers as well.”

Having just moved suburbs, this strikes a chord. I know the difference between a good and bad GP is huge. I know the quality of my auto-mechanic can affect my wallet in a big way. But I don’t know how to tell the difference by looking. I’d hate to be having to hire software engineers without knowledge of the field.

Hi Julian,
I think we’re in agreement, but just rounding off the rough edges here.

“I wonder though, what you call the field of software development that includes user experience design. (Also, people have already started dying from computer glitches. We passed that milestone decades ago.)”

I think (to some extent) there’s space for a vocation which doesn’t exist yet: As the architect is to the civil engineer is the something to the software engineer. I think that will come with time. Software engineering as a term has gotten an almost dangerous amount of breadth, so that maybe anyone could call themselves a software engineer? Not quite, but getting there.

As for death, I think it needs to be a particular kind of death. People in the army have died from software glitches, and to some extent that’s seen as socially acceptable. It’s when a software glitch causes, say, people in a building to suffocate to death because a central computer crashed and killed both the oxygen circulation and the lifts, that people will get serious about this sort of thing.

I’m not dismissing your career. Rather, I’m holding it up as having done it’s job! Compilers had challenges, they were overcome (some probably by you specifically!). You could do it again to gain an understanding of what the problem and solution was, but it’s not adding new value. Interestingly, if you look at programming languages built for the internet, you won’t find much. There’s PHP (which is… PHP), and .NET (but like you said, that’s mostly marketing, not really in the language), and erlang (but no one really uses it on the web). Maybe there’s room to create an “internet programming language”!

Google search is an excellent example of what I was thinking of. We’re on the same page.

For pointers, pass-by-reference, etc. I take your point. I was really only trying to say that you ought to understand the abstractions of your language, but if you don’t understand particular ones, it’s OK.

In C++, you specifically have to state that your member function is virtual. Overriding doesn’t work otherwise. In Java Overriding always works. It may be implemented in the same way, but java people don’t have to know it (I doubt many java people could draw up how overriding is implemented in the JVM).

I do see (and partially agree with) Michael’s sentiment. It’s sort of a Nintendo problem of people implementing E.T. might put people off gaming entirely — similarly crappy developers doing crappy web development might put people off computers entirely. This is a broader problem than the web, however, and the “one bad developer creates two new software development jobs a year” idea applies here.

For an auto mechanic it’s harder, but for a GP, at least in theory, you should be getting a consistent high quality due to the medical associations in place to ensure that nobody operates under the “doctor” title unless they have a certain minimum standard.