I tend to take a non trivial algorithm I know well and do my best to implement it in the new language.

Bingo. When I found myself with some time to spare and a burning need to figure out how OO Perl worked, I created Roff as an OO exercise in about a day. I learned a lot about how to implment a simple OO Perl program, and although I left myself room to expand it into a more complicated group of classes, never did get around to it.

This follows on from writing a version of roff in C sometime in February 1993 after I broke my leg and was without a computer for six weeks. In a frenzy of 14-hour days, I wrote a really pretty cool version that even did micro-justification on the dot matrix printer I'd rented.

The nice thing about roff is that you can just keep on adding features to it -- there's no limit to when it's done.

Which can be a good thing or a bad thing. :)