A tricky situation has arisen over a Linux patent (US: 5,995,745) obtained by Victor Yodaiken of Real-Time Linux. Although he intends that users of RTLinux, as well as Linux users who label their application as being compatible with RTLinux and who release them under GPL will not have to pay royalties, it is a narrow path that he is walking. Already there are questions as to what the situation will be for non-Linux free products like BSD and Hurd. Those who wish to use his patent in commercial systems will have to pay a royalty. Yodaiken is apparently working out details with Linux International and Linus Torvalds. But there is another issue, raised by Greg Aharonian in his Patnews: he calls the patent "of low quality", suggests that "this patent wasn't validly sought", and draws attention to the considerable inadequacy of the prior art. Yodaiken quotes just four other patents and a paper on MERT, a Unix time-sharing system described in 1978. But as Aharonian points out, IEEE started its annual symposia on real-time systems in 1979 and they have continued to the present day, making several hundred papers. Quite apart from these, there have been thousands of other papers on the subject, and hundreds of patents, all of which appear to add up to the inevitable conclusion that the patent was not examined properly by the highly inadequate US patent office. It seems rather unlikely that in all this work there is no other mention of what Yodaiken claims to be his invention: preemption in a general purpose operating system for real-time tasks - and after all, even VMS did this. The wider issue of software patents generally, and defensive patents for the Linux community, needs much more attention. The only solution would appear to be some legislation to require a public comment period when prior art could be submitted, or comments made about the novelty. This would go a long way to removing the need for expensive legal procedures to strike down patents that should not have been issued. ®