Action A-639-06: bug 29408

To implement the WGs decision on this I am proposing, unless anyone objects, to rewrite the paragraph on xs:decimal precision as follows:


<p>For <code>xs:decimal</code> values<phrase diff="add" at="D">, let <var>N</var> be the number of digits
            of precision supported by the implementation, and let <var>M</var> (<code>M &lt;= N</code>) be the minimum limit on the number of digits
            required for conformance (18 digits for XSD 1.0, 16 digits for XSD 1.1)</phrase>. Then for addition, subtraction, and multiplication
               operations, the returned result <rfc2119>should</rfc2119> be accurate to <var>N</var> digits of precision, and for division and modulus operations,
               the returned result <rfc2119>should</rfc2119> be accurate to at least <var>M</var> digits of precision. 
               The actual precision is <termref def="implementation-defined"/>. If the number
                    of digits in the mathematical result exceeds the number of digits that the implementation
                    retains for that operation, the result is truncated or rounded in an <termref def="implementation-defined"/> manner.</p>
            
            <note><p diff="add" at="D">This Recommendation does not specify whether <code>xs:decimal</code> operations are fixed point or floating point.
            In an implementation using floating point it is possible for very simple operations to require more digits of precision than
            are available; for example adding <code>1e100</code> to <code>1e-100</code> requires 200 digits of precision for an
               accurate representation of the result.</p></note>


Michael Kay
Saxonica