> This suggests serializing the Templates object (including the translet
> class(es)), which makes perfect sense.
Yes indeed.
> If this proves to be just as efficient as instanciating a
> translet class then I may be convinced.
Since the class should just be holding a string or two, it should be pretty
good.
I'm still not sure about the feasibility of creating a Templates object
that directly holds the bytecodes.
> or he/she may not prefer to use TrAX at all.
I certainly agree with this. Though it would be really nice if the
Templates interface worked well enough that TrAX could become the only main
public interface. The problem, in my mind, is that the extra indirection
of the Templates object makes it a bit clunky.
> XSLTC has been and should still be for a niche,
I don't think I agree with this. XSLTC in the long run should be suitable
for all runtime applications of XSLT that require high performance.
-scott
