RE: Wide Finder in XSLT --> deriving new requirements

> But I would for the processor to be able to do a good job without a hint. 

I think that to make the decision well, you need a lot of information about:

* the machine on which you are running

* the volumetric characteristics of the data you are processing

* what the user wants to optimize for - resource consumption or elapsed time
to completion?

Generally an optimization done without hints is ten times better than one
that has to be switched on by the user, because only a tiny fraction of
users will ever work out how to turn the knobs. (You can't even trust people
doing comparative benchmarks to get it right, if comparative benchmarks are
what drives you.) But when an "optimization" is capable of making things
worse when chosen injudiciously, I tend to be cautious.

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