XOM API for Processing XML with Java

[Janek Bogucki <janekdb AT yahoo.co.uk>]

I ran the same classloading benchmark with two versions of the the xom jar

    jar A: compressed, no index
    jar B: compressed, with index (META-INF/INDEX.LIST)

Note: the distributed jar is uncompressed; see previous posting.

Once again I think the motivation for using the index option for the jar was 
to improve class
loading time but I would be happy to learn otherwise.

This is the result of running ClassesLoader 20 times with each jar. Times are 
in milliseconds.
ClassesLoader was executed in a loop hence the pairings which are otherwise 
insignificant. The sum
of the times is at the bottom.

      jar A   jar B

        449     446
        436     450
        451     441
        442     454
        438     438
        449     444
        437     450
        444     439
        442     454
        439     445
        449     443
        396     450
        442     399
        445     446
        438     447
        444     443
        444     450
        443     439
        444     447
        440     448
                
        8812    8873

Adding an index to the jar is harmful in this situation. In fact indexes are 
intended to be added
to the root jar of a set of jars used by an applet. I doubt xom will ever be 
used as the root jar
of an applet. As it stands the xom jar is an uncompressed jar with an index 
is a curious beast
indeed, at least from an applet's point of view.

I suggest the removal of the index unless it serves a purpose other than 
class loading
optimization (which it fails to do).

-Janek


        
        
                
___________________________________________________________
Yahoo! Messenger - Communicate instantly..."Ping" 
your friends today! Download Messenger Now 
http://uk.messenger.yahoo.com/download/index.html