I agree programmers should be able to read and understand Knuth's books, but I don't think a CS degree is preparation. Would you think someone needs a CS degree if they already have a good understanding of the books?

My complaint with college is that it seems to be mostly fluff. If you show up and pay the bills then you'll get your degree.

* writing * accounting * organizational psychology * small amounts of business stuff * lots of EE, digital and analog * how to read and write research papers * how to research * how to prove theorems (important if you are inventing your own) * linear algebra, calculus, stats and probability, discrete algebra * physics I-III, plus 1 semester of experimental physics * chemistry (I,II, no organic) * a bit of philosophy * circuit design * numerical computation

It would be a Herculean task to take all that up in your spare time to the level I did at school (I still learn on my own time, so you have to catch up not just with what I did at school, but everything afterwards). It's 4+ years of opportunity to bump against really good minds. I went to professors and said "I want to learn x". (think of the course catalog as a suggestion!) That got me, among other things a co-authorship with a professor on an academic book. All of that has been relevant to my career except for the chemistry. Even with that, my first job could have used it if I had stuck around more, and it is useful to know if you want to be scientifically literate.

Or, you know, you can punch your time cards, do the bare minimum, and graduate with a pretty useless certificate.

College as voc tech to learn to manipulate the LAMP stack is probably a poor choice of time. If you want to be able to take a job to compute cancer statistics, program robots, build digital interface cards, perform computer vision, simulate the ocean, write aircraft wing simulations, model and research traffic flow, work on medical devices, you almost certainly need more then votech and/or self learning, IMO. I've done most in that list professionally, and friends of mine have done the rest (no rockets in that list, but I was offered a job to do that, turned it down).

There are multiple paths to life, but the chance to just think and learn for awhile is pretty incredible if you can afford it.

THIS IS THE BEST COMMENT YOU WILL READ TODAY, IF NOT THIS WEEK!!!!

Sorry for the all caps, but "truer words were never spoken".

So, anyone reading this who hasn't gone to college or who hasn't yet graduated, heed those words.

You will thank the poster perhaps 30 years later.

The year being over, I asked my friend if his brother is done with the books. He got back to me saying that his brother asked if I was in 1st year, and that he was surprised when he learned that I graduated and still studying.

I had a chance to talk with him yesterday (well, this morning at 4 AM), we were seven people at my friend's.

I said that not working harder than I did back in college is a mistake I am paying the price for today. I urged him to work harder than he is for he will later regret it with hot tears. I said that right now, he's in a cocoon where nothing is expected of him except to study, which he should do vigorously, relentlessly, and constantly. He has no other obligations or pressure other than to learn and become excellent. I said if I was asked to kill ten people to be able to go back a decade, I would kill an additional two to show how thankful I am for the bargain.

I tried to convey the message because he didn't fully grasp that time wasted is never, ever coming back. That once it's gone, it is forever. That it is the only thing we own only for its very duration, not a second more.

I hope he understood. There was a fleeting expression of realization after the chuckles that I hope will grow.

Thus, follow your heart- dwell not on what you did, nor on what you do, but listen to what you need. Long after, look back, and see, more clearly, what you need now.

> Would you think someone needs a CS degree if they already have a good understanding of the books?

To be a programmer? No. But, at least to me, programming related matters is but a small part of computer science. Which makes Knuth's books such a great resource: it goes into detail in that small, yet quite large, field of programming.

I guess what I'm saying is that an undergraduate CS degree is just a vocational degree. College does not prepare you to read TAoCP. If you want a solid CS foundation you have to self study.

I think my undergraduate CS degree contained exactly four courses that I would classify as pure vocational in nature: the first class everyone had to take was to learn to program in Delphi; a class on functional programming; a class on GUI programming; and a class on object oriented programming. After that, all classes were either:

- mathematics, from analysis, statistics, writing/reading proofs, to loads of discrete mathematics

- formal methods, from proving correctness of programs, modeling, to different methods to analyze complex systems

- theory, from automata theory, language theory, compiler theory, database theory, relational algebra, to complexity and computability

- odds and ends, such as ethics, philosophy, history

Then, each trimester, we also had an engineering project where we would have to apply theory. There were no vocational classes associated. If we had to, or wanted to, use a tool, language, system, we had to learn it by ourselves first. Nevertheless, these engineering projects can be seen as vocational in nature too, of course. Still, that makes for about one vocational class a year (of about 15), and three engineering projects a year.

So, when people talk about undergraduate computer science degrees, I always assume their experiences are not dissimilar to mine.

Even this single section within a single chapter could be used as the basis of an entire course.

The primary benefit of TAOCS is that it's self-contained and all written in the same voice/style. It's much higher quality than a lot of other texts on the subjects it covers, but a given chapter is not necessarily that best text for the given topic.

It would be really cool if there were a "TAOCS" elective offered every single semester, just for the sake of fun. I would love to teach a course like that. I think it would work really well for e.g., an honors program ("to get honors in the major, one option is to enroll in this course for 4 consecutive semesters")

More generally, I think the policy that courses should use TAOCS where a chaper exists for the topic at hand would certainly increase the quality of courses at non-top CS programs (certainly, it would have at mine).

But in general, there are often texts that do a better job at covering a topic. E.g., this is a really interesting chapter, but I think there are probably better introductory texts on SAT (entirely subjective, of course).

Which is to say, if I know the area well then I will probably not use TAOCS. But if I don't then TAOCS is a safe choice.

There are a lot more problems that need to be solved at higher levels these days and many of them aren't even in computer science. That's were I spend most of my research time - the coding/cs stuff is easy now.

The point I was trying to make was that Knuth is one of the granddaddies of this whole reality that we populate and that was my attempt to express my gratitude.

I'm infinitely curious about everything, but at some point in life you realise that you can't go too deep in all directions, because it's a fractal of knowledge out there, so some things you just have to trust to be true and move towards new grounds.

I have one of the first printings of the three first volumes but have not read all parts. I did do a deep dive in the floating point, as we at one point had to write that at MWC for an early compiler for the non-8087 class of computers. (Good work, Tim M.!)

Seibel: Have you read Knuth's, The Art5 of Computer Programming?

Zawinski: I haven't.

Check the book.

Not sure what book I have in mind, then. I know there was one where they focus for many interviewees was whether or not they read other people's code.

And, I should add that I don't see it as a failing that people don't read some literature. I do think it is more approachable than it often gets credit for. To the point that I wish I had tried to read it years ago. That said, people that are being obviously productive should remain so using whatever strategy they have used to this point. :)

I will add that learning algorithms in the way he presents them has been hugely successful in really understanding working with abstractions. Not just inventing the cheap abstractions of giving names to things; but really understanding how you can abstract some problems onto others and use common tools in solving them.

But I can't possibly allocate so much time reading these (long and complicated) books, because, well, all these problems are solved and implemented by people smarter than myself.

They're in the libraries, codecs, standards or embedded in the programming languages.

Problems I'm trying to solve are further and further away from that deep abstract level - and more in the realms of social psychology, perception, etc. Thanks to Knuth and others, I can move away from the abstract to the concrete, from the computer to the human mind, even if I couldn't possibly explain the fundamentals in such detail as he does.