Uncertainties about format-number

[David_Marston@lotus.com]

When I sent a copy of my message, originally to the XSL spec editors at
W3C, pointing out the difficulties of ascertaining conformant behavior for
format-number, I mainly wanted to raise the general issues. I agree with
Ken's thought that I have not yet said anything about how much of the
negative-number pattern is supposed to apply.

Here are the general issues as I see them.
1. The spec says that the various decimal-format attributes set values to
be applied like those of JDK 1.1 DecimalFormat, but it allows for divergent
interpretations of what result is expected. I think you should get the
result that was *intended* for JDK 1.1 as opposed to whatever JDK 1.1 might
actually produce. Worse yet, the intent of JDK 1.1 is not documented in any
public spec of which I am aware.

2. There are interactions among the decimal-format attributes that are not
documented. Most of the formatting characters use positional information,
so the numeric value has to be aligned with the pattern at one character to
establish positions for the rest.

The issue of the negative-number pattern is a problem with respect to both
of the above general issues.
.................David Marston