I've also decided to embrace my inner Don Quixote, for a few simple reasons:
If I can really build a kick ass killer-app webserver in scheme I will show the world; Google, Thoughtworks, and Amazon will all be falling all over themselves to hire me
I will end up building some useful libraries for Gambit Scheme
I'll be able to erase the sheer suffering caused by such monstrosities as PHP, J2EE, et all. (I pick on those 2 because I know them the best, but others are not exempt).
It's fun, and I am leaning more about HTTP, functional programming, threads, C programming, parsing and all sorts of great stuff.
Note that while points 1 and 3 have me firmly pointing my lance in the direction of the Netherlands (tilting at windmills, see?), points 2 and 4 are reasonable, observe:
Version 0.2 of my MySQL FFI for Gambit Scheme is available. The source is more documented, and there is FINALLY a semi-reasonable test suite. There is still a lot of work to be done, but it is getting to the point where there is at least a reasonable baseline.