It took Knuth the better part of a decade ...

In that decade, Knuth developed versions of TeX in Sail, Pascal, and finally even built his own "Web" programming system (Tangle/Weave) around the Pascal implementation. He also developed Metafont, and did three generations of Computer Modern typefaces. I know of at least one student of his whose Ph.D. topic was related to working on TeX (the hyphenation algorithms). Not to mention he carried on the rest of his academic work (although he did take sabbaticals). In typical Knuthian fashion, his analysis carried him from a relatively simple task (fix the poor printed quality of Volume 1) to a huge multi-year endeavor.

I know a great example for this, but I can't seem to find a link. I clearly recall a sidebar to the "Literate Programming" paper by Knuth that appeared in Comm. of ACM. Knuth developed a three- or four-page program to show how something would look in Web (actually, it was a little cheating, because what was published was actually the Weave output, not the original input text). Immediately following it was a short critique by Jon Bentley (of "Programming Pearls"). He developed a three-line shell script that accomplished the same thing. Granted, it wasn't publication-worthy, but it got the job done in a fraction of the time.

So while I agree with you that it will probably take more than the 15 years it took Knuth to duplicate what he did, I disagree that something "better" implies "bigger" -- quite the opposite, really. (Not to imply that I think developing a good typesetting program is easy or quick.)

If a certain Finnish student had believed that about operating systems, the computing landscape would look different today!