Re: more closed-world chatter

Marshall wrote:
> On May 11, 4:03 am, Jon Heggland <jon.heggl..._at_idi.ntnu.no> wrote:

>> Marshall wrote:
>>> If we are joining X with DUM, then there are necessarily no
>>> attributes in common, so we would not expect to perturb
>>> the attribute types from X.
>> So you won't consider each type of the result to be additionally
>> constrained by "and false", thus leading to the bottom type?

I'm not really convinced one way or the other myself. I'll read up on what D&D have to say on the matter, but the join-subtype thang sounds more reasonable than the AND FALSE-bottom one.

> I don't have the clarity on this issue yet that I need to.
> I'm used to thinking about traditional type systems
> but I'm trying to transition to thinking about more
> general constraint systems, and sometimes I confuse
> myself.

So you suggest a type system where types are sets, defined by constraints (or enumerations? You could formulate an enumeration as a constraint, I guess); and may have names (for convenience), but don't need them; subtypes/supertypes are implicit, based on set inclusion; and other types may be constructed/inferred/expressed as intersections, unions and differences between other types?

And the type of A in "T where A = 'J'" is the set { 'J' }, if the type of A in T includes 'J', and { } otherwise?

Sounds interesting, but undecidable, and not restrictive enough for my taste. :)

-- 
Jon