Fred Brooks vs. Dijkstra? (2015)

Fred Brooks vs. Dijkstra? (2015)(hmijailblog.blogspot.com)

35 points by hmijail 7 years ago | 15 comments

javajosh 7 years ago |

I believe we are still in the alchemical age of software application design. There are glimmers of enlightenment here and there that will help the transition to chemistry, but the transition hasn't really begun. Software done well today requires genius, but this should not be, just as basic chemistry labs don't require it.

Djikstra was complaining about these alchemical methods, and rightly so. The sad and dangerous thing is how much we have built on the shaky edifice of alchemy, and how much we blame ourselves for failure, and how we pride ourselves when we master some local improvement. (Even worse, the pride of finding some new local improvement).

hmijail 7 years ago | |

(Author here)

I agree. There's much too much "art", religion and cargo cultism in software development, and much too little science/engineering.

And it's hard to even say whether we're improving.

mcguire 7 years ago |

"He talks about using OOP to remove "accidental difficulty" from the process, “allowing the designer to express the essence of the design without having to express large amounts of syntactic material that add no information content”…

"…but he stops at OOP? Makes me think that he was not very informed about functional programming, given that OOP is (1) trivial to add to functional programming languages, (2) dissed by all the greats as uninteresting – or worse."

Here's my take: OO may be a decent way of managing complexity, but it introduces a (noticably high) complexity floor that negates many of the benefits. It works very well for some things, specifically windowed guis, and less well for almost everything else. Rule of thumb: if you are modeling the "problem domain" and not the "solution domain," you are doing it wrong.

However, for better or worse, every programmer in the last 30 years has been trained on OOP; most simply cannot conceive of not using a "class" keyword. On the other hand, almost no one does OOP itself anymore. Objects, classes, are treated as a module system, which is noticably missing, or at least very limited, in most programing environments.

All problems have some level of essential complexity, although often it is lower than both the floor demanded of both the programing environment and the programmer's self esteem. (It's difficult for many to admit that what they are being asked to do is trivial, so they look for ways to make it more difficult, including introducing bizarre monstrosities into their environment.)

On the other hand, exactly no one wants to grapple in hand to hand combat with any moderately complex problem (and many problems do have sudden, steep, complexities) as Dijkstra would have preferred. Or, some problems cannot be solved in the necessary resource constraints, including the brain power of the programmer. As a result, more complexity is pushed down stream---and modern systems have very long tributaries.

So, what can ya' do?

mpweiher 7 years ago | |

> [OO] introduces a (noticably high) complexity

Hmm...I'd say Smalltalk is about as simple as things get.

For anything more than a few functions/methods, you want some sort of scoping/module concept, and I think you're right that that's one way classes are used. Might as well be classes as they are lightweight, provide a convenient metaphor and other benefits.

I would say that true OO programming is rare. Most I've seen in industry is scoped procedural with little to no polymorphism. Or paraphrasing Alan Kay: "Before you dismiss OO, you should at least try it once".

marcosdumay 7 years ago | | |

Smalltalk is simple. But Smalltalk doesn't help making your programs simpler - at least not as much as a powerful FP language.

OOP is a convenient tool for cutting your state into smaller, independent pieces. But that can only go as far as the pieces can be independent, what for most practical domains is not very much. It also does not bring any other big benefit bundled with it, what looks to me like the EDJ complaint is all about.

contravariant 7 years ago | |

Assuming well defined problems have a single solution, how exactly is the solution domain different from the problem domain?

mcguire 7 years ago | | |

No real problem has a single solution.

olau 7 years ago |

Clickbait. The post does not contrast Fred Brooks with Dijkstra.

(Instead it presents some complaints from Dijkstra - on overblown marketing material mostly. And presents some of Fred Brooks well-known points - apparently the author doesn't quite agree, or perhaps hadn't grasped the depth of Brooks' thinking.)

hmijail 7 years ago | |

(Author here)

The post quotes both Brooks and Dijkstra; the first managed development of a system and got famous, the second saw the system's actual failures and reported his criticism on it.

So, care to explain what exactly is it that you find lacking? My experience with people who cite "failure to grasp the thinking" has been rather with "architecty" types that can't get past handwaving to make a point, so anything concrete would be appreciated.

For example: you say Dijkstra complains "on overblown marketing material mostly". Citation, please? Sounds like you mean that the marketing overpromised and underdelivered, but to me it looks like Dijkstra was pretty clear about the whole set of ideas being poppycock even if fully delivered; he says that it's fundamentally the wrong way to do things, and does so in a rather smug "told you so" mode.

... unless we want to argue that maybe Dijkstra didn't "grasp the depth of Brooks' thinking" either, nor the functionality of that thinking's actual, first hand implementation?