Unified Numbers

From OWL
Jump to: navigation, search

This is a summary of Alan Ruttenberg's objection to the OWL WG's decision to require disjointness of primitive number types. The issue boils down to the question of whether OWL users are better served by having a floating point literal 3 and an integer literal 3 be interpreted to be the same value, or different values. Alan's position is that because OWL is a language for modeling real-world domains, and not an implementation language, it would be more useful to OWL users (i.e. lead to better models with fewer errors) for OWL to provide a single type of numbers, with integers and floats as subtypes, rather than artificially distinguish between implementation types, making integer and float (for example) disjoint.

The desirability of a single number type was not in dispute. In July 2008, as a possible resolution to its issue 126, Boris Motik prepared a proposal for a treatment of numbers, based on a paper he coauthored with Ian Horrocks. The proposal made decimal and floating point both be subtypes of a real number type. At its face-to-face in July 2008, the working group felt there was consensus on the proposal, and the design was instituted, eventually reaching last call in December 2008.

However, precipitated by LC comments 22 and 24, the WG revisited its design at the Feb 2009 F2F, as both comments expressed the concern that RIF number types were disjoint and therefore incompatible with OWL. Most of the WG agreed that the compatibility problem was serious, leading to a resolution soon thereafter to reverse its decision and define that number types should be disjoint. Alan registered an objection at this time as the sole no vote against this resolution.

In response to concerns about RIF compatibility, Alan consulted Jos de Bruijn, a member of the RIF WG; Alan's summary argues that RIF's reliance on existing XPath libraries, which seems to be the source of the difficulty, creates as many problems for RIF's relation to OWL as it solves (see the last two sentences of the email for summary). On the related issue of XML Schema Datatypes 1.1 draft compatibility, Alan sought an opinion from its editor, Dave Peterson. According to Alan Peterson's response supports the notion that XSD type disjointness does not necessarily apply to OWL.

In addition to the issue of RIF compatibility, Boris Motik observed that comparison of floats to integers was difficult. Alan responded with a citation to a published paper describing a technique for efficiently comparing floats to integers.

(Prepared by Jonathan Rees with help from Alan, 31 May 2009)