Amit Agarwal points out that a couple of court rulings make it illegal to deep link to the copyrighted files. The rulings are about audio and video files, but the same logic can be applied to any non-HTML resource.
Though Internet is all about linking to resources, deep linking can cause use of bandwidth without a visit to the web site. Deep linking is good, but it is beneficial to the resource owner only if the link is to a HTML page as it encourages visits. If I provide link on my blog to download a file residing on other web site, my readers can directly download it without having to visit the web site. Though it might be convenient for my readers, for the file owner it is a cost of the bandwidth without the benefit of recognition through visits or page views. In case there is a HTML page referencing the resource, a link to the HTML page can be more beneficial can a deep link to the resource.
Many, including yours truly, assume that anything on the Internet is linkable. And that the content owner can take enough precautions and the onus is on him/her to avoid others from deep-linking. These rulings might mean that the onus is on the one who is linking to verify the copyright before linking. A little tedious, but acceptable because I know a lot of content owners who do not know how to workaround these things. Of course this applies only to copyrighted resources.