Re: model theory of error [Yes, proposing one!]

On Jan 12, 2008, at 12:29 AM, Michael Kifer wrote:
[snip]
> It is not easy to "see" something like that. One needs to check,  
> which is
> very time consuming.  My crystal ball says that the chances of  
> breakage are
> over 50%.

As another data point, Christian, I did think about this a couple of  
times, and I ran through two or three ways (including yours and a  
"parameterizable" model theory) to the point of starting email  
because I thought I had a simple, obviously workable compromise. They  
broke hard almost as soon as I started writing them out.

What's wrong, from your perspective, in saying that the default  
semantics are (a), but implementations might not conform to the  
semantics for formulae with built-ins when there is an error?  
Implementations should document whether they conform and this is a  
place where RIF may not be perfectly faithful *by default*.

Cheers,
Bijan.

Received on Saturday, 12 January 2008 00:39:06 UTC