search

w3c / sdw-sosa-ssn

Repository of the Spatial Data on the Web Working Group for the SOSA/SSN vocabulary
7 stars 5 forks source link

Remove OWL cardinality restrictions #216

Closed dr-shorthair closed 1 month ago

dr-shorthair commented 3 months ago
  1. take the Ontology outside DL
  2. raises various problems ... https://github.com/w3c/sdw-sosa-ssn/pull/206#issuecomment-2010877514 https://github.com/w3c/sdw-sosa-ssn/pull/209#issuecomment-2007293991

probably more.

Mostly I added them where I though they made sense, but cardinalities are probably better managed at the application level using SHACL, rather in the Ontology layer where they affect reasoning and can cause difficult-to-control inconsistencies.

ldesousa commented 3 months ago

I would like to follow on the comment by @rgcmme. Constraints modelled alone at application level (e.g. with SHACL) would inevitably lead to misinterpretations of the ontology.

An example I am currently dealing with is the relation between Observation and Procedure. Can/should an observation refer to more than one procedure? In OMS the restriction is set by the class diagram itself, wheres in SOSA/SSN the sosa:usedProcedure object property has no cardinality indication. Therefore a SOSA observation with two different procedures is valid, whereas it is not so in OMS/STA.

I don't have the experience with reasoning to sense the troubles cardinalities could imply. Therefore I admit those difficulties to eventually prevent their specification, but in that case I would prefer to have them absent altogether.

oldskeptic commented 3 months ago

I would suggest leaving cardinality restrictions in SOSA. They are unlikely to break things on the reasoner side and if that's the SOSA design, that's what the ontology should say.

OWL and SHACL tend to draw different people but this level of restriction isn't problematic IMHO.

dr-shorthair commented 1 month ago

I have found it necessary to remind people that OWL cardinality restrictions are not validation constraints. The OWA still seems to be a step too far for some smart users.

AFAICT the one that really matters crossing over from OWA to CWA is where cardinality=1 - that means that if there is more than one, then either they are identical or there is an inconsistency.

But I will close this issue now as no-one else wants to deprecate cardinalities.