Re: Outcomes of RIF telecon 16 Dec

Great summary, Chris.
Let me comment on constraints once again so that we could continue the
discussion by email. This might help move this issue along, since Hassan was
not at the last telecon and I won't be at the next one (and possible the
one after that as well).

Chris Welty <cawelty@gmail.com> wrote:
> 
> 
> Although we did not pass any formal resolution today, we feel that the
> WG is close to consensus on a couple of key issues:
> 

.........

> =Constraints=
> 
> We did not have time for much more discussion, and Hassan was not present. Michael 
> presented his position that now that he has tried to work constraints into the CORE, 
> he does not see how it can be done while still maintaining the design principles that 
> the CORE be a core.  No one on the call claimed to both "understand Hassan's 
> proposal" and "support it", and in fact there was only one of the latter.  No one 
> seems to be objecting to it yet, but there seemed to be consensus that the pros and 
> cons of constraints in the CORE were not well-understood enough at this point and 
> need further discussion; and that constraints should be left out of the first public 
> working draft as a consequence (without prejudice towards future decisions wrt later 
> WDs).
> 
> -C&C


My main problems with putting CLP in the core are these (the 2nd one being
a show-stopped as far as I can see):

1. There doesn't seem to be any particular role for constraints in the core
   unless we include some concrete domains there and mandate support for
   some constraint solvers over these domains.
   In the absence of this, CLP would be just syntactic baggage in the core. 
   Not clear why it shouldn't be in the dialects that require this scheme.

2. By a previous decision, all dialects must extend the core both
   syntactically and semantically.
   It seems that there is a general agreement (if not a decision) that
   eventually we will have a full FOL dialect and an LP dialect 
   based on the most popular semantics for negation as failure (well
   founded or stable or both).

   - It is not clear to me how CLP plays out in the FOL setting and whether
     it makes sense there in the first place. In FOL one should be able to
     apply arbitrary connectives and quantifiers to constraints, and it is
     not clear how this works with CLP. (Of course, we could treat constraint
     predicates as builtins, but for that we don't need CLP.)

   - For the LP dialect with NAF (well-founded or stable), CLP hasn't
     really been developed theoretically of practically.  The most
     comprehensive LP system that supports well-founded negation is XSB
     AFAIK. I asked David Warren about XSB's support for constraints.  He
     replied that "one must be very-very careful" with them :-).  He
     confirmed that this is basically for experts who know what they are
     doing (by deeply rooted intuition :-). The theory of the stuff is
     mainly lacking.

   Given all this I cannot see how we can have CLP in the core and be
   consistent with the extensible design decision.
   On the other hand, there is no problem with developing dialects where
   CLP plays important role.


	--michael

Received on Wednesday, 17 January 2007 22:30:30 UTC