[r6rs-discuss] [Formal] Allow compilers to reject obvious violations

[Matthias Felleisen]

On Feb 25, 2007, at 12:07 AM, r6rs-discuss-request at lists.r6rs.org wrote:

> Matthias Felleisen wrote:
>>   2. Naturally I don't reject type systems per se but I think that  
>> a serious
>>      language definition shouldn't introduce such systems without  
>> specifying
>>      them. Otherwise a language/implementation will appear  
>> whimsical to
>>      programmers.
>
> The current draft already mandates hundreds of runtime
> exceptions whose whimsical purpose is to make programs
> that violate the requirements of the R6RS less likely
> to run to completion.  Why should that kind of whimsy
> be limited to run time?

This is not true. The report comes with a formal and executable  
specification of the operational semantics of R6R Scheme. I can  
determine whether or not a program's behavior is within the  
parameters of this specification.

What I am asking for is an equally precise specification of the "type  
system." That's all. (And I would be happy with stylized English that  
spells out the analysis that compilers are expected to perform.)

-- Matthias