[Seaside] Re: those manic algorithmic questions

[Tim Rowledge]

Well, it looks like it might be a simpler problem than I feared (thank 
Cthulu) that needs solving. The idea is not to have to generate such 
complicated multiple choice questions at each view request, but to 
produce algorithmic fill-in-the-blank type questions when the question 
is added to a test.

I think I've demonstrated that the initial scary idea is actually bad; 
imagine taking a test where you get asked a bit of simple arithmetic 
dressed up as an accounting question (all accounting is just dressed up 
trivial arithmetic), say
Q1. what is 3 + 4?
	[text box] )you type in 7
and then next time you go to that same question it is different (being a 
magical dynamic question), say
Q1. what is 6 + 23?
--ack! what should appear in the text box where you had previousy typed 
7? 29, on the grounds that you got it right previously? Eek.

You probably see the problem.

So, now I can actually create a more sensible system where there are 
template questions that get instantiated when aded to a test. Once 
added, they are fixed, like quails in aspic. I might even implement this 
as trivial classes with simple methods to implement the algorithm and a 
number of collections of possible parameters and explicatory strings. 
I'm sure there must be a more elegant answer though!

tim

tim@sumeru.stanford.edu