This is an archived snapshot of W3C's public bugzilla bug tracker, decommissioned in April 2019. Please see the home page for more details.
The value space of float and double should be aligned with IEEE and with the XML Query, XPath, and XSLT specifications and should have two distinct signed zeroes. Historical background: the first edition of XML Schema 1.0 had two distinct zeroes, but there was pushback over the consequence that for purposes of checking minimum and maximum values on bounded types +0 was greater than -0. To respond to these concerns, the WG issued an erratum which made them a single value. This in turn distressed some observers who felt that alignment with IEEE and QT required distinct signed zeroes. Eventually, the WG concluded that the correct approach is to have distinct values, but make bounds checking depend not solely on identity but instead on numeric equality: positive and negative zeroes are to be distinct values which compare equal in bounds checking. Since this is the behavior of conforming IEEE arithmetic libraries, it should be unsurprising to most people, although some observers have already deplored the distinction thus made between identity and (numeric) equality.
A wording proposal for this issue is included in the omnibus package of 31 August 2005. http://www.w3.org/XML/Group/2004/06/xmlschema-2/datatypes.omnibus.20050831.html
New proposal sent to WG per decisions taken at telcon of 2005-10-21: http://www.w3.org/XML/Group/2004/06/xmlschema- 2/datatypes.b1917.20051028.html#float
I'm slighty uncertain of the current status of this item. The wording included in the omnibus proposal of 31 August was approved by the WG in Edinburgh, which would seem to have closed the item. I do not know why a revised proposal was thought necessary (and wonder whether the proposal sent to the WG in October was actually for a different issue related to float and double). In any case, the wording approved in Edinburgh was integrated into the status quo document 8 Dec 2005.
(In reply to comment #3) > I do not know why a revised proposal was > thought necessary (and wonder whether the proposal sent to the WG > in October was actually for a different issue related to > float and double). Precisely; it was to close out the RQ-21 (lex reps and mappings, and reformat to add a value space subsection) rewrite. Which same was subsequently approved and is awaiting integration of WG-directed amendments and then will be incorporated int SQ.
(In reply to comment #4) > (In reply to comment #3) > > I do not know why a revised proposal was > > thought necessary (and wonder whether the proposal sent to the WG > > in October was actually for a different issue related to > > float and double). > > Precisely; it was to close out the RQ-21 (lex reps and mappings, and > reformat to add a value space subsection) rewrite. > > Which same was subsequently approved and is awaiting integration of > WG-directed amendments and then will be incorporated int SQ. No note indicating that the material described was in fact incorporated into the SQ document, but one editor (MSM) marked the material RESOLVED FIXED after that comment was posted, so I trust that they were so incorporated. Michael, please check, and mark this bug CLOSED if they were in fact incorporated.
(In reply to comment #5) > No note indicating that the material described was in fact incorporated into > the SQ document, but one editor (MSM) marked the material RESOLVED FIXED after > that comment was posted, so I trust that they were so incorporated. Michael, > please check, and mark this bug CLOSED if they were in fact incorporated. Oops. ' Twas not this bug that was referred to as needing a rewrite, it was "the RQ-21 rewrite" (which happens to be bug 1907), which has been put in the SQ document and CLOSED. Accordingly, I retract the request for checking, and am herewith marking this bug CLOSED. Mea culpa.