> Thanks a lot Mike for your observations.. I propose
> this code to solve the attribute ordering problem you
> cited (for both the documents).
> 
> <xsl:for-each select="$doc1//@*">
>   <xsl:sort select="." />
> 
> i.e. adding a xsl:sort instruction in the for-each
> loop. This shall solve this problem!

No, you need to sort on the attribute name. Two attributes can have the same
value in which case their respective order is still unpredictable.

> *I feel my stylesheet caters to a large subset of
> Canonical XML definition (I guess about 70-90%). Can
> you please comment on my this claim? *

I've no idea what the claim means. Are you saying that your algorithm will
give the same answer as a canonicalized comparison 70%-90% of the time? If
not, what are you saying?

Michael Kay
http://www.saxonica.com/