RE: TR/xquery-operators/#func-trace

> quote. The location and format of the trace data set are 
> *implementation defined*. unquote.
> 
> So we may have this 'debug' information
> a) Added to the serialised output.
> b) Output to an error file
> c) Output to a terminal screen.
>  And any one could be selected by an XSLT implementation?
> Which screws portability nicely?
> 
> 
> For XSLT I'd request the output 'location' (If I understand that term
> correctly)
> be defined. I.e. to follow the content of xsl:message.
> 

The format and destination of the output of xsl:message have always been
implementation-defined. I would expect most XSLT implementors to send the
output of fn:trace() to the same destination as the output of xsl:message,
but I don't see any need to constrain them to do so. Once we make things
implementation-defined, I think we should make them completely
implementation-defined.

Michael Kay

Received on Monday, 12 May 2003 22:40:43 UTC