[Imps] Test cases for parsing spec

[Sam Ruby]

Henri Sivonen wrote:
> 
> This may sound like bikeshedding, but that's not the simplest thing that 
> could possibly work.

While that's a fair statement, I do believe that James's previous 
statement (namely "since it can be parsed easily, we will be able to 
convert it to a better format in the future.") very much applies here.

My instinct is that having each test specify the full set of output 
tokens is both a burden and a maintenance headache, James's opinion is 
clearly otherwise; neither one of us is sure, but in the end the data 
will speak for itself.  Similarly, perhaps it would be easier if each 
test was separate (it certainly makes it easier to run exactly one 
test), of perhaps that's not important.

And, if you recall, I also previously voted against 'noise characters', 
of which the current syntax has many.

All in all, I'd suggest that we let James go for a few days and the 
reassess.  Given this readily parsable syntax (in JavaScript and Python, 
at least), converting to a new syntax should only take minutes.

- Sam Ruby