See also: IRC log
<csma> PROPOSED: accept the minutes from MArch 9
<csma> RESOLVED: accept the minutes from March 9
<ChrisW> close action-993
<trackbot> ACTION-993 Contact jderoo about DTB support or get from implementation report closed
<ChrisW> close action-992
<trackbot> ACTION-992 Contact changhai about DTB support closed
<ChrisW> close action-991
<trackbot> ACTION-991 Contact chime about DTB support closed
<ChrisW> close action-988
<trackbot> ACTION-988 Add note to BLD, Core about plain literals closed
<trackbot> ACTION-942 Draft reponse to TK3 (Thomas Krekeler) closed
<trackbot> ACTION-938 Draft reply to public comment AR4 closed
<trackbot> ACTION-824 Review PRD operational and model=theoretic conditions are = closed
<trackbot> ACTION-990 Contact Mike Dean and ask which DTB type are supported closed
<trackbot> ACTION-989 Contact mdean for list of datatypes closed
sandro: need to respond to public comment from Paul Gearon on 2009-11-5
<ChrisW> ACTION: sandro reply to paul gearon [recorded in http://www.w3.org/2010/03/16-rif-minutes.html#action01]
<trackbot> Created ACTION-994 - Reply to paul gearon [on Sandro Hawke - due 2010-03-23].
<ChrisW> ACTION: chris to update public comments list [recorded in http://www.w3.org/2010/03/16-rif-minutes.html#action02]
<trackbot> Created ACTION-995 - Update public comments list [on Christopher Welty - due 2010-03-23].
<ChrisW> Sandro summarizes issue from sparql wg
<ChrisW> ... want to use sparql to use results of rif inference
sandro: issue is we want to use
sparql to query
... sparql works on graphs
... to apply to rif, query a graph that imports rif
so sparql would have rif import
<AxelPolleres> SPARQL can define what rules to refer to, only if we have a handle from the graph to the rules.
sandro: invites people to help,
expects sparql group will have doc to review in a couple
... rdf syntax for rif has no semantics, must use axel's import to give meaning to the rules
<AxelPolleres> we thought that defining such mechanism using a rif: owned URI would be better than a sparql: uri, since importing rif rules from RDF... and there meaingin to a triples/graphs entailed by the resp. rif-rdf-combination, may be of wider interest than SPARQL only.
csma: is rif entailment in rdf graph?
... may have several import profiles to address some corner cases
csma: why use rif namespace?
sandro: seems natural if in rdf and want to include rif, to use rif namespace
<AxelPolleres> rif-owned URI would be better than a sparql: uri, since importing rif rules from RDF... and thereby meaning to a triples/graphs entailed by the resp. rif-rdf-combination, may be of wider interest than SPARQL only.
<AxelPolleres> +1 to sandro
<AxelPolleres> like owl:imports
csma: but the import stmt is in the sparql spec, so I expect to see sparql URI
<AxelPolleres> we assumed that such mechanism is potentially of wider use than sparql only
<AxelPolleres> ... thus more natural in rif:
sandro: could be in any namespace
(e.g. make up a new one) but easier for users to have
... we should review the working draft and see if it is clear
chris: seconds Sandro
<sandro> chris: let SPARQL proceed using RIF namespace; we'll review future drafts to make sure they use it reasonably well.
<sandro> (general agreement)
csma: missing 1 core impl
chris: we have a schema validator
sandro: but there may be syntax that is not schema validatable
<sandro> schema can't tell if you're in Core or in a dialect that allows free variables.
csma: would have to check no equality in head, only ground lists, etc
<sandro> (but I would argue that such a dialect is evil --- all extensions must include some syntactic marker of that extension)
csma: is a schema-validator enough?
sandro: I think so
<sandro> sandro: Although there may be pathological dialects which cannot be distinguished from core, that's the failure of the dialect.
sandro: if you cannot distinguish a dialect from Core using schema validation, you have a poorly designed dialect
csma: we consider a Core schema validator sufficient to meet the exit criteria
IRIS implements all of DTB
<csma> # ( ?arg1; pred:is-literal-boolean( ?arg1 ) )
<csma> # ( ?arg1; pred:is-literal-PlainLiteral ( ?arg1 ) )
<csma> # ( ?arg1; pred:is-literal-XMLLiteral ( ?arg1 ) )
<csma> # ( ?arg1; pred:is-literal-not-boolean( ?arg1 ) )
<csma> # ( ?arg1; pred:is-literal-not-PlainLiteral ( ?arg1 ) )
csma: we are missing a small number of DTB builtins to have 2 impls of each
<csma> # ( ?arg1; pred:is-literal-not-XMLLiteral ( ?arg1 ) )
sandro: there could be other bits of other specs (PRD, etc.) that are not complete
csma: some of the larger PRD test cases have no XML
we are missing some test coverage
csma: and no impl (ILog, Oracle) passes all the PRD tests
sandro: need to set the "bar". Highest is we test all impls. We aren't going to do that. More practical is to lower the bar.
chris: e.g. OWL had some test cases that were never implemented.
<sandro> see http://www.w3.org/2003/08/owl-systems/test-results-out#bycount and you see for OWL 1 there were zero test cases with nothing passing them and 8 test cases with only one system passing them.
csma: is it ok to exit without
good impls but with promising implementations in the
... e.g. current impls may require some "fixup" by hand
sandro: I think that is probably enough
for BLD we have IRIS, Vampire prime, and Silk implementations
scribe: also Eye but some confusion about its lack of XML parser
chris: unioned DTB support from all reported implementations, RIFle, OBR, Eye to determine short list of missing builtins
csma: how can Eye be a consumer if it does not consume RIF XML?
chris: also fuxi has no parser
... must be using some other "presentation syntax"
csma: RDF support is the biggest issue
sandro: need rdf parser that can
... plus need to pass import test cases
chris: RIFle has some RDF import support
csma: but RIFle is only a validator, cannot pass entailment tests for import
sandro: will try to implement RDF import soon
<Harold> Implementations update DONE.