The copyright itself is not a major problem. The problem is with "related" IP issues, particularly in light of the on-going W3C IP policy debate over royalty-free vs. reasonable and non-discriminatory (RAND) licensing terms. W3C apparently would not be content even with us putting the DAML+OIL documents in the public domain. Also, journals will accept Lucent's copyright release, but W3C requires use of its own form.
Peter's understanding is that only the submitter has to sign the IP release, while every author may need to transfer copyright.
We want to keep this information out of the query language primarily to keep the model theory for query clean and simple. Peter referred to this as a "Platonic version of querying", to which specific interfaces can be defined.
KQML and OKBC support such additional information regarding packaging of results and resource consumption.
There was a discussion about plurality, e.g. creating a bag within the query to contain a set of results, rather than depending upon external mechanisms to return multiple sets of results. While theoretically clean, this may be cumbersome for developers.
The representation of the query, additional information, and results remains an open issue. See Pat's previous message.
Ian expressed the view that
is primarily intended to define the language itself,
and should generally not be used in applications
because of its limited semantics.
There was a discussion about "anonymous objects" (e.g. inferred objects, such as the unnamed father associated with a Person) and whether queries should return them. There was general agreement that a reasoner shouldn't export internal names, only constants in the KB or underlying representation, however Pat advised us not to underestimate the "evil ingenuity of logic programmers".
We revisited the issues
(primarily monotonicity vs. practicality)
with supporting queries such as
Pat noted that we should distinguish between querying databases vs. knowledge bases.
$Id: 2001-10-23.html,v 1.5 2001/11/06 18:07:14 mdean Exp $