I really hate it when I come across a reference to a paper or a standard and I am not able to read it because I must buy it or be a member of a subscription plan at prices that are atrocious by the standards of developing nations. In my line of interest (Computer Science), the biggest culprits are the ACM Digital Library and ISO. Fortunately, most of the times I can work around this limitation by using Google with a query like '"<title of the paper>" filetype:pdf' or by taking the help of a friend. I really wish researchers use arXiv or an equivalent service more often. The whole point of publishing papers is to help advance your field, not to retard it, right? (Apparently, the situation is still bearable in Computer Science but is unbearably ridiculous in basic sciences like Physics and Mathematics; hence the genesis of arXiv.) If I ever publish a useful book, an article or a paper, I would try my best to make it publicly available for all.