Well, it is fiendishly hard to have a quantitative measure of quality in software. From your own example, a bridge can be precisely qualified in this sense: it will take a maximum of so much load and it can withstand so much external forces at such and such points. One can even use simulations and scaled models to test the designs. The external factors are fewer and well documented (traffic, wind speed, temperatures, water flow etc.). So no wonder, the bar is higher. On the other hand, software almost has uncountable states/external factors and one can only test so much. There is always a combination of factors which developers and tester never thought of, that bring the whole thing down. Of course, this does not excuse rolling out of software with known bugs and issues.

Firefox 3.0 still wins on complex script support on Linux (due to better integration with pango than Firefox 2.0?). I don't know about other complex scripts, but for Devanagari (variants of this script are used for almost all north Indian languages, e.g., Hindi) Opera 9.5 fail to render ligatures on Linux correctly. I submitted a bug report way back when it was in beta, still not fixed in the final release. Come on, it is 2008 and a browser cannot render scripts used by 1 billion+ people.