He's not exactly comparing software to netscape or win 3.11 though, he's comparing version N of some software to version N-1 or N-2 and noticing that they're getting worse from release to release. Given the rate of new releases the complexity shouldn't be increasing that rapidly between releases so I'm not convinced that is the cause per se. I have to agree with the conclusion from the article, testing was more rigorous in the past than it is now. Both because there was less surface area to test back then and because time-to-market pressures were less due to the longer windows between releases.