[linux-elitists] 32 essential computer books?

Jim Thompson jim@netgate.com
Thu Dec 18 13:19:31 PST 2003



Ben Woodard writes:
> On Thu, 2003-12-18 at 10:02, Jim Thompson wrote:
> > Tanner Lovelace writes:
> > > Don Marti wrote:
> > > > Well, I just looked at the shelf by my desk and 32 books fit
> > > > on it.  Strangely enough, I also found out a way to send books
> > > > to the Linux User Group of Iraq, which needs books for its
> > > > library, at the US book rate.
> > > > 
> > > > So what would be the top 32 books to send them?
> > 
> > The Art of Computer Programming, Volumes 1-3, Donald Knuth.
> 
> I think that these are fine books but I'd have to argue with you
> regarding them.  I think it is important for any programmer to have good
> algorithms and data structures experience but these books are not the
> way to get it. Plus they are terribly out of date not necessarily with
> regard to the information that they have but with regards to the place
> that the information has in the wider realm of computer programming.

Oh wow.  ReallY?

I agree that MIX is baroque, but its due to be replaced with a RISC
machine (MMIX).

All three existing volumes continue to be translated into interesting
"3rd world" languages (Polish, Russian, Chinese (technically not 3rd
world, I know), so <someone> must be getting some use out of them.

Vol4 (if it ever actually does arrive has work on combinatorial
algorithms

And ya know, he has generated some very bug-free code (backing same
with his personal funds), wich is more than anyone can say about glibc
or its ilk.

> I think that the day has passed when we want people doing their own
> implementation of X,Y, or Z data structure or algorithm. I think for
> maintainability sake, and to foster faster code development we should be
> strongly suggesting to the underlings that they use already extant
> libraries like glib with reasonably bug free and optimized versions of
> many of the fundamental building blocks. Later on when you are writing
> something very specific that needs to be carefully optimized, e.g.
> google's caching algorithm, you can delve deeper than thing that are
> handily available in freely available libraries.

but until one understands the math (the 'why') algorithm 'a' is faster
(or uses less space) than algorithm 'b', then its somewhat for naught,
eh?

And this is what TAOCP is about.

> The only one that I would really recommend is volume 2 the one on
> semi-numerical algorithms but that is really for people who want to do
> computation not necessarily people who want to use computers. But
> knowing first hand what a lot of these numerical algorithms are actually
> used for, I can honestly say that at this point in world history and
> with the current political and law enforcement climate as it is, I
> wouldn't feel exceedingly comfortable sending Knuth TAOCP Vol. 2 to
> Iraqis for much the same reason that I think it would be unwise to be
> seen sending them textbooks on how to build supercomputing clusters.
> A little later on I think that it would be fine but right now, I think
> that it might throw up a red flag that might hinder later efforts.

Books are free (as in speech), 1st amendment protected, etc.  (Geez,
this *is* "linux-elitists", no?)

Jim

-- 
"Speed, it seems to me, provides the one genuinely modern pleasure."
			-- Aldous Huxley (1894 - 1963)




More information about the linux-elitists mailing list