Searching the primary literature on the Internet leaves a lot to be desired. For example, nothing puts the brakes on a search faster than finding a reference to a key paper, but being unable to download it due to the access policies of the hosting journal.
Enter ChemRefer. The site works very simply: type in a keyword and get a list of papers matching it somewhere in the article. Click on one of the search results links, and you'll get the entire article as a PDF file.
What makes this remarkable is that most of ChemRefer's holdings appear to come from copyrighted, subscription-only journals. For example, searching for "taxol" produces links to the full PDFs for papers published in Tetrahedron Letters., Org. Letters, and J. Med. Chem., among others.
It's not clear whether ChemRefer is legally or illegally redistributing these articles. The results are not comprehensive, indicating some form of pre-selection. The journals in question are known for their tough copyright stance and high prices, so there may be a problem regardless of how the material was originally obtained. ChemRefer's founder, William James Griffiths was interviewed on Reactive Reports about his service.
Regardless of the legal status of its holdings, ChemRefer offers an intriguing window into another world. What wonders might result if independent Internet services could mash-up the primary literature at will?