Re: model theory of error

Michael Kifer wrote:
> 
> So, if it is a compliance clause, then we pushed the problem to the next level.
> Does it mean that I can now go back to work on the document?

Yes, I think that the benefits and drawbacks of the three options are 
clear enough and we should now make a decision.

I have no sense of where the WG is leaning, though?

*Sandro*, would it be possible to set up a quick WBS poll to see what 
kind of resolution would have the best chances? As I understand it, the 
options are:

a. Add a special constant, e.g. rif:error, and a distinguished elements 
_|_ in the domain of interpretation of any semantic structure, such that 
all semantic structures will be required to interpret rif:error as _|_; 
define builtin functions so that they will return rif:error whenever 
their arguments are of the wrong type; and say that a predicate, 
p(a,b,c,...), is false if any of its arguments evaluates to _|_ in the 
interpretation;

b. Same as (a) as regards builtin functions. This option introduces a 
new truth value, called E (error) such that ~E = E, E/\F=F, E/\T=E, 
E\/F=E, E\/T=T. Then we can say that p(a,b,c,...) has truth value E if 
at least one of the args is _|_;

c. Support builtins (and evaluated functions and predicates in general) 
syntactically, and semantically within their domains of definition, but 
leave their semantics on error (and the semantics of predicates where 
one arguments is an evaluated function with an out-of-domain argument) 
implementation dependent.

Whether or not silent processing of such cases should be forbidden is a 
different issue. It depends on how difficult that requirement would be 
to implement, as I see it.

Christian

Received on Thursday, 10 January 2008 19:53:19 UTC