Calculus texts are almost universally
1) large
2) expensive
3) written by "math people", for "math people"
4) a whole lot like every other calculus text
Let's address these issues one at a time. First, size is an issue because many people never use the whole text, which is usually designed to cover everything any institution would want in 1.5 to 2 years of calculus. This leads to unnecessary tree killing and the second objection, unnecessary expense. You should be able to buy just as much calculus book as you need, but this is usually not the case. The third objection really applies to most mathematics texts. The book is already big enough, and thus there isn't room to provide enough explanation for the average human to "get it". And the books seem to be written with not only the primary, but almost exclusive goal of accuracy. That is, they appear to be written strictly for the purpose of meeting with approval from other mathematicians scanning the text for errors. That may make them accurate, but it doesn't do much to help them be readable to other kinds of humans.
The last objection is that, as many calculus texts as there are in print, you will find that most of them cover something close to exactly the same material in the same order with the same proofs and nearly identical problem sets.
Is all of this necessary?
Thankfully, the answer is no. These facts are simply evidence that textbook publishing as currently practiced lives in the more-or-less-recent past.
Let's try to get a perspective that's a bit more modern. What is a calculus text? Usually a collection of words, equations, and illustrations. In past decades, if thirty people in a class needed to be able to independently access a collection of words, equations, and illustrations, it would, presumably, require thirty books. That's a bit anachronistic now--although there is still a place for books, it seems pretty silly to create so many hard copies of the same information when it could be stored on a single machine and served out to student's computers over a network.
But does this take care of the above objections? The answer is, "It depends how you do it." The size issue is the easy one. The information can be stored quite compactly on some kind of reusable storage device. But what about expense? You can be sure that the same publishing companies that want a hundred bucks a pop for the paper version will be charging a hefty license fee for use of the electronic version, right? And since the same publishing companies will be writing them, shouldn't we expect the content to remain about the same? So what about this "by math people, for math people" stuff and the fact that everyone covers about the same material in about the same way? What makes you think that will change just because it's stored electronically?
The answer is, they won't change _just_ because the text is stored electronically. As long as the same companies are in control of the texts, we wouldn't expect much to change, short of the possible saving of a few trees. But that's where the statement "it depends how you do it" comes in. Something fundamental _did_ change when we began thinking of the text as being stored on a computer connected to a network. Because that network could be the Internet. And that text could be Open Text. Think Linux.
He gave it away, with the source code, and let the world do it for him. Sort of. What Linus Torvalds did was to release his code under the constraints of the GNU Public License-- instead of a copyright, which forbids others to use the content without permission, the GNU Public License is a "copyleft", which allows you free use of the code, but forbids _you_ to copyright your derivative work. In other words, you can change it, but you have to make your changes available to the rest of the world, for free, just like it was given to you.
The result of this was that many, many people got the code, played with whatever part of it interested them, and contributed their improvements. Linus watched over the whole project and had final say over what was added to the "core" code. But before anything got added, it was seen and tested by people all over the world (because the Internet and the copyleft made it easy to distribute) who improved it and sent in their suggestions to the author (because the Internet made _that_ easy to do, and there's no motivation to keep your copylefted code a secret--you can't sell it, so the best thing you can do is tell others about it and make the whole project better, adding a little fame to your name along the way). It took a while, but within a few years Linux was really, really good. And now you're probably wondering "Why is he still talking about a computer operating system?".
Or maybe you're not. Maybe you already see where this is going. A copylefted calculus text on the Internet could easily spawn the same sort of peer-reviewed, massively parallel process that could shape _it_ into the best-of-breed calculus text. And one of the neat side benefits of this is that this great end product would be available to anyone in the world, for free.
Suppose instead that you could fire off an email to the maintainer of that part of the text, meanwhile posting your own version to a newsgroup for anyone to grab who wanted it.
The only question is whether there are enough other people out there like me that have explainer's itch. If you don't believe there are, you've never read Usenet news.
We also need to figure out a painless but reliable way to have changes and submissions come in and get added. That shouldn't be too hard, either, but it's critical that it work well, or the whole thing is doomed.
Another important consideration is the question of "Who will be Linus?". Someone has to be accepted by the community as having the final say on what goes in, or we need a Perlescent way of doing it by committee. In either case, though, it's critical to get someone who people have a reason to trust.