First off -- I applaud your use of open-note exams. That is the ONLY real-world way to learn and demonstrate knowledge. There is almost never a situation in the professional world where one must solve a problem with absolutely no references (and it would be stupid to do so on a production system -- when solving a critical problem, why risk everything based on what you *think* is right, when you can verify against documentation; at least if something breaks, you can point to the incorrect docs...)
Some people can simply memorize anything they look at, while others struggle at this. A proper exam should be designed to test one's ability to demonstrate processes: exams should give you all the information you need, but the questions should be designed such that only someone who has invested prior effort in practice and learning will be able to solve the questions in the allotted time.
For less-concrete subjects such as the arts, I'm not so sure how this can be accomplished. However this is a trivial design decision for exams in maths, sciences, programming, and engineering.
Furthermore, I think any physics or math exam that requires a complex calculator really has a wrong approach. Assuming everyone at this level has already demonstrated their ability to perform arithmetic several times over, the calculator should only be there to free them from making mistakes on the menial number crunching (addition, subtraction, multiplication, division, squares, squareroots, proper value of e,Pi, etc...). The exam should test for core concepts: ideas where you simply must understand the knowledge through prior practice and learning.
Sadly, I think many professors fall back on rote-memorization exams just because they can't be bothered to design proper exams each semester. These types often teach straight from the textbook-provided lesson plans, and then wonder why students cheat...
But honestly -- an exam is but one facet of demonstrating proficiency in a subject. Personally, I think projects & labs the best way: sure one can cheat, but it's easy to determine who has spent time polishing a proper unique lab report. In this respect, open-ended projects are the best, as the room for creativity limits the possibility for undetectable cheating, and lets the students show their enthusiasm for the subject. If you're really worried about cheating, a lab-practical may even be a legitimate tool: it's pretty damn hard to make stuff up as you go while you've got a one-person audience of the professor.
Short answer: let them use basic scientific calculators, the textbook, their notes, and a dictionary; design your tests so that students have all the resources they need, but don't have enough time to learn-as-they-go during the exam.
"Never memorize something that you can look up." --Albert Einstein