RE: XSLT 2.0 processor signalling error, for type erro

> Just out of curiosity, if I write say
> 
> int i = "abc";
> 
> in Java
> 
> I would get a compilation failure (even if the variable is not used).

Java is a statically-typed language.
> 
> Most of the time, XSLT compiles the program and executes it 
> in one pass. Can we say XSLT is a interpreted language, like 
> say JavaScript?

Compiled vs interpreted is a matter of the processor implementation, nothing
to do with the language itself. And it's a pretty fuzzy distinction these
days, because most processors actually compile down to some intermediate
representation which is then interpreted by some kind of virtual machine:
the question is how close that representation is to machine code. JIT
compilation makes it even fuzzier.

A clearer distinction is between strongly-typed and weakly-typed languages,
or (almost the same) statically-typed and dynamically-typed. XSLT is
essentially dynamically-typed (meaning type-checking isn't done until
runtime) except that there's provision for optional optimistic static typing
- which means you can report a compile-time error if you find a construct
that's guaranteed to fail if executed at run-time (as in your example), but
not (like Java) if there's merely a possibility that it might fail.

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