Scientific software is worthy of being cited on its own, without reference to a paper in which the software is described, and citations of the software should be tracked in the same way one would for a standard journal article.
Some authors of software feel they need to publish an article that describes the software, and they then ask users to cite the paper rather than the software. This suggests that software is a second-class product, not worthy of being cited directly. I understand the reason they do this: administrators, in counting beans, tend to value print publications more than other products, and some administrators might not even realize that Science Citation Index tabulates software citations directly. However, by asking users to cite the print publication rather than the software directly, programmers are helping to devalue their own product. If the software authors don’t fight for a belief in the value of software, who will?
Fortunately, many journals now allow one to cite software directly, and some have a special style for doing so. However, some publishers (including two with whom I have dealt recently) require one to cite the “data accessed” if the software is downloaded from a website. In so requiring, the publishers are confusing the means of obtaining the product from the product itself. Most if not all software has a version number. If the version number is included in the citation, why include the date the software was downloaded? That’s like requiring us to include in citations of printed papers the date that the journal issue arrived in the mail, or the date a book arrived from Amazon.com. How silly.