Every time I think "today is the day when I'm actually going to work through some of TAOCP" I remember that I'd have to learn MMIX.
Knuth is obviously one-of-a-kind, but for me personally learning a new architecture requires more brain space than I have to spare, and pretty much guarantees I'll never read it. Which makes me a bit sad.
I know that at the time, C looked like it might pass by like Pascal did before it, but in retrospect I think C would have aged far better and made the book far more approachable.
It's a bit ironic, because one of the main justifications for using a machine language was that high-level languages go in and out of style. But Knuth did not avoid this himself, as MIX (a CISC-like architecture) was obsoleted and had to be replaced with MMIX (a RISC-like architecture)!
You don't actually need to use MIX; my course used TAOCP and skipped most of the assembly-level stuff, and everyone did fine just psuedo-coding things. I've never written a single line of MIX.
I'm looking forward to this. I can't claim to have rigorously done every exercise, but I did read the books closely enough to know what was in there if I ever need to, say, write a garbage collector.
The real reason to read them, though, is because roughly every third page will have something you just have to play with, like the neat things in the part on circular lists (sorry to be vague, it's been a few years). Buy it: there's so much in there you're guaranteed to find something fascinating.
I've also read volumes I-III closely enough to do the easiest problems, and would like to find the time to go back and perhaps study some of the more interesting parts in more depth.
Knuth gets accused of being abstruse, but compared to most math books I've read, he's very clear. Moreover, he has a genuine unpretentious fascination with the things about which he's writing, and that enthusiasm comes through in his writing.
For that reason I'll be buying and enjoying the fourth volume, although I'd like to wait until all three "sub" volumes are published so I can get them in a box.
Unfortunately, though, I don't think Knuth would be a good reference for garbage collection anymore. That particular problem (and several others covered in the books) have moved so quickly that Knuth's descriptions haven't aged well. His most advanced example of GC was a crude mark-and-sweep system, which would be horribly archaic if implemented today.
PS: I had occasion a while back to implement a random number generator that produced numbers from a normal distribution. It was a lot of fun to crack Volume II open and use it as my reference (after poking around a bit to make sure the information wasn't obsolete).
Hopefully, if I am writing a garbage collector from scratch, it won't be in a context where performance matters. :) I thought the best part of that section was that the garbage collector ran in constant space, which makes sense as a requirement but isn't something I had thought about before.
> Knuth gets accused of being abstruse, but compared to most math books I've read, he's very clear. Moreover, he has a genuine unpretentious fascination with the things about which he's writing, and that enthusiasm comes through in his writing.
I totally agree. I guess the complaints are because where other books just say "see? get it?", TAOCP keeps going, down to measuring the complexity (not the order of complexity, actually counting how many times each instruction gets executed, which I think is why he uses MIX instead of a higher-level language).
. . . Man, now I want to drop Elements of Computing Systems and go back and read volume I again.
I just hope he lives long enough to complete the rest of the book(s) he has planned. This thing took so long that it's, like, the Duke Nukem Forever of CS books. :-(
Duke Nukem Forever was on preorder in 2001, too. The status of DNF is "never" until people are actually playing the game...just because it's on preorder again doesn't mean it's actually going to come out.
The obvious connection, of course, is that the developers of Duke Nukem Forever needed to read Volume 4A (in fascicle form, at least) in order to actually complete the development of the game. ;-)
If anyone is planning preordering via this link, here is how to do it to get the lowest possible price:
Don't apply the coupon code. You actually get it for a lower price if you just add it to your cart and checkout as normal. There is an extra deal going on right now that lowers the price another 5% on orders over $55. Applying the coupon cancels this extra deal.
You end up getting the book for $149.99 vs $162.49 if you use the coupon. Great deal, and 50 dollars lower than amazon's preorder price.
Meanwhile, the electrical engineers have been waiting for Art of Electronics v3 for a decade... "its coming out real soon, its being edited right now!"
It's amusing that they're releasing a box set for Volumes 1-4A, although I suppose it took longer to get Volume 4A to press than the entirety of the Harry Potter series (and they released a boxed set after pretty much every volume of HP).
That's by far the best price I've seen for the collected volumes though.
I wouldn't want to be the publisher betting on a 73 year old with a history of missing deadlines hitting one in nine years time.
And then, from the site:
---
After Volume 5 has been completed, I will revise Volumes 1--3 again to bring them up to date. In particular, the new material for those volumes that has been issued in beta-test fascicles will be incorporated at that time.
Then I will publish a ``reader's digest'' edition of Volumes 1--5, condensing the most important material into a single book.
And after Volumes 1--5 are done, God willing, I plan to publish Volume 6 (the theory of context-free languages) and Volume 7 (Compiler techniques), but only if the things I want to say about those topics are still relevant and still haven't been said. Volumes 1--5 represent the central core of computer programming for sequential machines; the subjects of Volumes 6 and 7 are important but more specialized.
---
Which would all be wonderful, but seems possibly overoptimistic. Sadly.
The current errata for volume 4A has 2 items: 1) he left off the left quotation mark around an equation in one place, 2) one entry in the index should be changed.
So people shouldn’t worry too much yet about the book being full of errors. :)
Color me surprised, I didn't know that InformIT was the publisher now. I guess I gotta get out more. I think I bought mine in the 80's (paid $43.95 for vol 1, list price).
Thanks for your helpful comment. But it looks rather like Addison-Wesley Professional exists only as a brand name now, and as a publisher they seem to have been fully absorbed, and their former offices gone some time ago by what I now gather.
Well, in the publishing world, a "brand name" is known as an imprint, and many major publishers fall into that category. For example, Penguin and Viking are also imprints of Pearson. Random House, Doubleday, Knopf, Bantam, Ballantine, Pantheon, and Vintage are all imprints of Bertelsmann. And so on...
And I apologize if my original response came across as snarky-- I was aiming for jovial.
No problem, I guess I am concerned about names like AW becoming not just acquired but absorbed in such a way that makes me wonder about the future. When I'm interested in a book, one of the things I look at is the publisher - if it is one of the names I respect, then they probably took the time to make sure the book isn't a waste of time.
No doubt staff is drastically reduced in these acquisitions. How does a high-quality name avoid falling apart? Do they start shoveling crap to compete?
The available titles in computing have certainly vastly increased since I started out (for that I am thankful), but for many publishers the quality of editing is very poor.
I think I have been putting this off far too long but... do you guys think one could study TAOCP alone, without studying at MIT or Berkeley? I have been working with computers since more than 18 years as a programmer and sys architect but my studies never included those awesome books like TAOCP.
I am not very fluent in maths, so now I am worried if it is at all understandable to study it on my own? Maybe I am completely off in my assumptions here.. I would appreciate some feedback!
Oh and, any other equally fundamental IT/computers/programming MUST_READS you can recommend?
One of the volumes (I think the subtitle was "Sorting and Searching") was used in my undergrad CS program, and I definitely didn't go to MIT or Berkeley.
It's comprehensible, though it's definitely not written in an "algorithms for dummys" style. Check it out from a library and give it a try... you might be surprised.
I don't have a computer science background and I recently read TAOCP and Introduction to Algorithms by Cormen et al for an interview with google. I failed the interview but got through the books finding them very interesting.
Reading TAOCP was a worthwhile experience but I feel the Introduction to Algorithms is of more practical use because it covers a lot more ground (at less depth) and is a much easier read.
If you allocate the time, and have the desire, you can approach TAOCP in any manner you want and get something out of it. I read the 3 volumes cover to cover, including reading the exercises, but not really doing them. I did gloss over some of the really long hard algorithm analyses. At the beginning of the first volume is a refresher on all the maths Knuth uses in the books (a little over 100 pages).
How so? I am sure that there might be some who would prefer to read a work as difficult (and intricately typeset) as TAOCP on a screen instead of paper, but I imagine they'd be a minority.
Despite what some may say TAOCP is a very readable series of books. But they do require being read in hardcopy, at a table, with a pen(cil) and notepad.
This isn't something you read on your Kindle while taking the Subway uptown.
No offense again, but all of the replies to my comment are equally ludicrous. If it were digitally available, you could search and index it, text mine it, distribute it to people in poorer countries, carry it with you when you move a lot. These are all the obvious advantages of ebooks, and nothing about TAOCP makes it especially necessary for a digital version. To make an educated guess, the problem must be some part of Knuth's attitude to life - he spent a major chunk of his life designing a system to make things you type in from a computer look good on paper.
Edit: to clarify, I'm not saying it shouldn't be on paper - freedom of choice! - I'm saying it should also be available digitally. And to the person who said that TAOCP is only a collector's item, you're completely ludicrous.
Knuth is obviously one-of-a-kind, but for me personally learning a new architecture requires more brain space than I have to spare, and pretty much guarantees I'll never read it. Which makes me a bit sad.
I know that at the time, C looked like it might pass by like Pascal did before it, but in retrospect I think C would have aged far better and made the book far more approachable.
It's a bit ironic, because one of the main justifications for using a machine language was that high-level languages go in and out of style. But Knuth did not avoid this himself, as MIX (a CISC-like architecture) was obsoleted and had to be replaced with MMIX (a RISC-like architecture)!