Re: Start discussion

On Mon, Mar 22, 2010 at 9:57 AM, Eric Prud'hommeaux <eric@w3.org> wrote:

> * Ezzat, Ahmed <Ahmed.Ezzat@hp.com> [2010-03-22 04:49+0000]
> >
> > Hi Juan,
> >
> > We have tasks for the use case.  I agree that I do not see enough
> discussion on the distribution list. It was agreed on we need the use case
> completed before diving deeper in the mapping language. This Tuesday let us
> discuss what is left on the use case. Our highest priority is to finalize
> what the team will be delivering sometime in April - higher priority than
> the semantics of the language.
> >
> > I agree for using Datalog in expressing the semantics of the mapping
> language; we should discuss that in the group. If I remember correctly, Andy
> Seaborne used Datalog in expressing the semantics of some SPARQL language
> constructs in the SPARQL WG...
> >
> > Lee, Independent of which approach you use, you need to validate the
> semantics of the mew language.  Advantage of Datalog, as it is based on
> logic, it is more expressive than relational algebra. Below is few pages
> about Datalog.
>
> In order to ground these rules on our goal of RDF, I've started a wiki page
> to capture the differences between representations in relational, datalog,
> rdf (SPARQL) and RIF:
>
>  http://www.w3.org/2001/sw/rdb2rdf/wiki/LogicalEquivalences
>
> The fact that datalog is recursive means that we can define mappings which
> can't be executed in an SQL query, or reallized as an SQL view. For
> instance, if NthLine (see the defn of SecondLine in the above page) were
> defined as having some recursive rule, like:
>
>  NthLine(emp, man) :- SecondLine(emp, inter) AND NthLine(inter, man)
>
> we would not be able to represent that in a relational form. This does not
> immediately disqualify such an expressivity, but it will seriously reduce
> the number of conformant implementations.
>

That is why we use non-recursive datalog per Perez et al :)


>
> > Regards,
> >
> > Ahmed
> >
> >
> >
> > -----Original Message-----
> > From: public-rdb2rdf-wg-request@w3.org [mailto:
> public-rdb2rdf-wg-request@w3.org] On Behalf Of Lee Feigenbaum
> > Sent: Sunday, March 21, 2010 5:51 PM
> > To: Juan Sequeda
> > Cc: public-rdb2rdf-wg@w3.org
> > Subject: Re: Start discussion
> >
> > On 3/21/2010 8:26 PM, Juan Sequeda wrote:
> > > Hi Everybody,
> > >
> > > There has been no discussion at all on the list, and that honestly
> > > worries me.
> >
> > I share this concern, which is one reason that I decided to join the
> > group and try to add what I can.
> >
> > > I know that we need to have a Use Case documents, but it is not
> > > completely clear to me what else we need to turn in and by when.
> > >
> > > One issue that I personally feel than needs to be settled is the
> > > semantics of the language. One we have this defined, it is just a
> matter
> > > of deciding on what is the syntax. I don't think we have made much
> > > progress on this issue. I have proposed to develop the semantics of the
> > > mapping language in datalog. I'd be up for working on this in
> > > conjunction with Marcelo Arenas and Dan Miranker.
> >
> > Ideally, I'd hope that the semantics of the language follow from the
> > requirements which follow from the use cases.
> >
> > As far as datalog, I know next to nothing about it, so my questions are
> > probably naive. Is there a datalog standard that we will be able to
> > normatively reference if the semantics of the language are given in
> > datalog?
> >
> > Also, I'm not sure if separating the syntax & semantics is definitely an
> > easy thing to do. I know that there are people on the group who advocate
> > that the mapping language be based on SPARQL CONSTRUCT queries combined
> > with a default mapping from the relational model to the RDF model. In
> > this case, I imagine it would make more sense to lean on the SPARQL
> > algebra/semantics?
> >
> > Lee
> >
> > > What else should we be having discussions on? The clock is ticking.
> > >
> > >
> > > Juan Sequeda
> > > +1-575-SEQ-UEDA
> > > www.juansequeda.com <http://www.juansequeda.com>
> >
> >
>
>
>
> --
> -ericP
>