Document first, implement later

Lately, in terms of »research« I've been working on a draft of the InforML manual. It is one-half specification, one half tutorial. The general idea being that if I have thought through everything well enough to explain how to program in the language, I probably understand it well enough to know exactly what needs to be done in implementing it (or rather extending/modifying AspectML to become InforML). So far this has worked really well, as every time I sit down to rewrite the bits of my tutorial, I keep coming with things to put on my list of design questions.

Another part of the exercise has been to focus on trimming out anything in the language that isn't actually necessary to complete my dissertation. However, if the feature is already in AspectML I'll probably leave it in unless its interactions with the extensions for InforML are not well understood and/or would require sitting down and doing some involved proofs to verify.

I think I'm converging and hope to have a kind of complete tutorial and specification ready by Friday, but we'll see.

Leave a Comment