- From: Michael Smith <msmith@clarkparsia.com>
- Date: Thu, 10 Jul 2008 09:28:06 -0400
- To: Rob Shearer <rob.shearer@comlab.ox.ac.uk>
- Cc: public-owl-wg@w3.org
On Thu, 2008-07-10 at 11:24 +0100, Rob Shearer wrote: > > Further, I have concern that because the lexical space allows > > arbitrary precision and due to alternative, inconsistent rounding > > implementations (permitted by XSD 1.1 [2], see 3.3.5.2) the lexical to value-space > > mapping could differ between implementations. Thus, two reasoners with > > xsd conformant lexical to value mapping algorithms could correctly > > produce different entailments. (I don't have access to IEEE-754 to > > look into the details here, I'm basing my statements on [2]). > > I've just been looking through IEEE-754. Although it does allow for > some differences in rounding implementations, it places (quite strict) > limits on the possible error introduced. My understanding is that this > means it is possible to encode any floating-point number in decimal to > sufficient precision that any IEEE-754-compliant translation back to > floating point results in exactly the same value. > Assuming that my understanding is correct, this means that it is > perfectly safe to serialize floats as `xsd:float` within an OWL file > without loss of precision: although there do exist `xsd:float` strings > whose exact values are not precisely defined, users always have > unambiguous representations available to them. Yes, I believe if we start with an IEEE-754 float value, it will go lexical to value space and back without loss. I was concerned about the large number of possible constants (e.g., "0.01") for which there is not a directly corresponding value in the float value space. [2] indicates to me that for some those, there are alternative lexical to value space mappings, which could cause problems. > So I recommend sticking with the semantics for lexical representations > defined in the XSchema spec. [2] says, "since IEEE allows some variation in rounding of values, processors conforming to this specification may exhibit some variation in their ·lexical mappings·. " It seems that unless we were to constrain the set of permissible float constants to only those which are mapped into the value space losslessly, then we have a problem. -- Mike Smith Clark & Parsia [2] http://www.w3.org/TR/xmlschema11-2/#float
Received on Thursday, 10 July 2008 13:28:53 UTC