Hacker News Reader

Comment by amichail | Hacker News Reader

Probably more of a status symbol for most people.

How many of you actually read these books carefully?

I've read parts of vol. 2, especially about random number generation. I found it in a library book sale for a quarter. THAT made my day. :)

I've found a suspiciously high percentage of the people who have totally! read the Art of Computer Programming are unable to answer even basic questions about MIX, by the way.

Well, they're difficult books to read to begin with. Knuth himself would be surprised if someone has read and understood them all. At the same time, I find they're hyped up a bit (like SICP). There are plenty of other really good algorithms books IMHO.

http://folklore.org/StoryView.py?story=Close_Encounters_of_t...

(Knuth denies this ever happened.)

At http://news.ycombinator.com/item?id=1266502 someone cites a particular occasion on which Knuth supposedly denied the story. As I remarked then, "full of shit" seems quite an unknuth expression.

That's pretty reasonable with MIX, given that it's a terrible language that was obsolete before it was invented and that all the algorithms are described in English as well.

No, like they'd never heard of it.

That's because there is a big difference between reading and understanding.

Reading can just mean reading word by word, symbol by symbol then putting the book on the shelf. That ends up happening a lot with material that has a lot of formulas and theorems. It could take days for some to understand just a couple of pages.

So basically I not surprised at all that you've met so many people who've read the volumes, but didn't all the subtleties.

I think reading the volumes is the next step in status show-off step after buying and putting them on the shelf.

Also, as many have pointed that MIX is just too obscure. Not knowing much about doesn't really indicate a lack of theoretical knowledge. I would personally skipped over those chapters.

I was more thinking along the lines of, "Oh yeah, I've read them all. Wait, MIX? What's that?"

I've seen several equally dense, informative CS books (such as Dick Grune's _Parsing Techniques: A Practical Guide_), but Knuth has a specific sort of cachet among xkcd-quoting geek hipsters, so people namedropping him, make a pretense of having read them all the way through, etc.

Volume 2 is my favorite as well. I think that was the first time I'd been exposed to Benford's Law (http://en.wikipedia.org/wiki/Benfords_law).

I keep them around as a reminder that Computer Science is math. Being able to reason and think mathematically is an important skill for any programmer.

A person can go pretty far doing web work and using scripting languages, but when you get into serious programming (heavily concurrent, resource-critical, real-time, etc) you need to understand the fundamentals.

Having Knuth up on the shelf reminds me to take a second and say, "Hey, think this through some more. Knuth probably figured it out already."

1. Me, at least, and I'm absolutely certain I'm far from unique.

2. There's nothing wrong or status-symbol-y about using books of this sort for reference, and reading only the bits one needs.

What's wrong with less cryptic books like these?

http://www.amazon.com/Algorithm-Design-Jon-Kleinberg/dp/0321...

http://www.amazon.com/Introduction-Algorithms-Third-Thomas-C...

Depth.

I have read CLRS (the second book in your list); it's really great as an introductory text as it spans a lot of areas. For example, sorting/searching is a chapter in CLRS, while it's a volume in itself in TAOCP (The Art Of Computer Programming). Also, CLRS is written in the style of a textbook, rather than a reference and a newbie would find it easier to follow. And, TAOCP is more of an in-depth reference to practically everything that has been researched in that field so far.

How many people need this level of depth?

A lot more than you would think - there are plenty of programmers out there that can't tell you the tradeoffs between the different approaches, and they can get away with it, because it appears to work (see thread synchronization, cross-site scripting, data-base injection, null pointers, pointer overflow, etc, etc) until somebody demonstrates that it doesn't.

thread synchronization, cross-site scripting, data-base injection, null pointers, pointer overflow

Does TAOCP cover any of those?

I would expect it to have plenty of algorithms to implement mutexes with, as well as parallel sort algorithms.

There are also going to be a lot of pointer based tricks that you don't get unless you really understand pointers.

A rather superficial status symbol, I'd say: owning a book does not equate to knowing or understanding its contents.

I purchased my copies individually (back when I was a student and I could ill afford them) and, ignoring Knuth's suggesting algorithm for reading the series, read the "interesting" sections quite carefully. I still use them for reference and occasionally find time to re-read parts of it.

Pretty ineffective status symbol, too. How many people whom you might want to impress are going to see exactly what's on your bookshelves, and know enough to recognize Knuth as impressive? I'll hazard a guess that most people for whom those are true also know enough to be more influenced by other more meaningful things than what books you happen to have on your shelves.

(I bet some people do buy Knuth as a sort of status signal to themselves, though.)

I wonder if heavily thumbed annotated editions cost more?

I read and understood volumes 1 to 3 front-to-back. I do not think it is that hard to understand, compared to e.g. complex analysis or number theory. Didn't work hard enough on some the exercises, but certainly understood all the simpler ones. I would not recommend reading this to every (would-be) programmer or computer scientist, though. Especially for those wanting to program on 1GHz+ CPUs with 1GB+ RAM, there are things that are more immediately useful. Also, even for the low-level stuff the content of the books is getting seriously dated (e.g. due to the absence of multi-threaded algorithms)

I have to admit I haven't read the books, lower head in shame. It was too expensive when I was a poor student. Interestingly years later at work there were some Romanian exchange students doing summer internship. When they left, we bought gifts for them. A coworker and I bought the 3rd Volume to give to one of them. That really made her day.

I use them as references probably more often than any other books I own. That said, I use the internets an order of magnitude more.

I read the first book and didn't find it particularly good. You don't learn a lot from them. You do learn a lot of irrelevant details (like how his machine language works).

The first three, in detail. I even tried and failed to do many of the higher exercises :-(

I used it as an algorithm reference for memory allocation.

Reply on news.ycombinator.com