[yummyfajitas]

Quality isn't the issue for most subjects. There are plenty of free and cheap textbooks with perfectly fine quality.

For example, when I took thermodynamics in college (in 2001), I used a thermo book from the early 70's which cost me $8 on eBay (as opposed to the modern version which cost $100). I also skipped purchasing many books entirely, and just used free materials online, including this newfangled site called wikipedia.

Agency costs and are the issue. The people assigning books don't have to pay the cost of using them, so they have no incentive to select based on price. Further, it's very difficult for students to use substitutes - if the professor assigns problems 1-10 from the book, it doesn't matter how good the material in your cheaper substitute is.

If you want to crack this nut, focus on problem sets, not textbooks.

[frak_your_couch]

Well, I do disagree on the quality aspect. I'd like to see a high quality Calc I and II textbook that's free, coherent, and not aimed at MIT students. My wife is a faculty member and if you can find me one, I'll pass it along (they're constantly looking for one). It's just hard to not choose Stewart..it's a solid textbook for a broad range of students, but it's expensive as shit. A lot of professors are getting that this (if they don't, I make damned sure I mention it when we socialize with them) is a barrier for a lot of students.

Anyway, you're absolutely 100% correct about problem sets. Good problem sets go a VERY long way to making a good textbook.

[ghshephard]

It's been 15 years since I took Calculus I and II, but looking over: http://www.khanacademy.org/#Calculus

Is there _anything_ in Stewart that isn't covered as well by Khan? You also get advantage of the video tutorial on the topic.

I'm waiting for that point in time when we reach a tipping point and Higher Education Institutions don't just delegate everything to Khan for a lot of their topics.

My Calculus Course (151) at SFU took place in Images Theater at 8:30 in the morning and had 450 Students listening to a lecture on Calculus. The ratio was approximately: 1/8 of the students were bored. 1/8 of the students were lost, 1/8 of the students were attentive, 1/8 of the students were asleep, and 1/2 of the students didn't show up. (8:30 AM!)

I see little need for Stewart in the Face of Khan's Video Collection on Calculus + a bit of wikipedia (Seriously - check out their section on Derivatives and compare it to Stewart)

I will agree with you on the Problem Sets - and perhaps that's the missing ingredient in Khan - He needs to more fully flesh out his test bank - but http://www.khanacademy.org/exercisedashboard is moving ahead nicely. (And, in my mind, is far superior to any static test bank)

[yummyfajitas]

Problem set quality is not the issue. If it were, I would donate a week or two of my time and build a problem set generator.

The issue is that you need the same problem sets as the instructor. If the instructor uses Stewart and homework is graded, you get a 0 on HW. This is why you can't even go back 1 edition of the book, and why the legacy publishers print a new edition every 2 years with minimal changes beyond tweaking the problems.

A startup idea I never pursued, but one I think has a lot of merit: build an online problem cross referencer. It provides a) free problems to any instructors who want to use them and b) crossreferences these problems with existing textbooks (that way students who don't want to buy the book don't have to).

I'm pretty sure that Int[ 1/Sqrt[3+5x^2], {x}] can't be copyrighted.

[pixcavator]

Int[ 1/Sqrt[3+5x^2], {x}] can't be copyrighted but a problem set that consists only of exercises as routine as this would be almost worthless. A cross referencer is a neat idea though.

[smanek]

http://tutorial.math.lamar.edu/ seems good (after 30 seconds of googling, and 5 minutes of skimming).