Issue1370

In olden times we decided to create our own id()s rumor has it due to 
heavy usage of System.identityHashCode affecting performance

I suspect that's no longer the case on modern day JVMs (but we'd need to 
confirm that). Using systemIdentityHash code would simplify our silly id 
implementation and ease debugging when using Java debugger/profiler 
tools

I thought that was the reason anyway, I'm not so sure from the commit 
message of:

http://fisheye3.atlassian.com/changelog/jython/?cs=2343

Note that regrtest passes just fine w/ identityHashCode (quickest way to 
try is have IdImpl.java_obj_id call it)

pedronis mentions the reason for it here 
http://markmail.org/message/fiuxxvjkdqhdtd4e#query:+page:1+mid:cmnq44kzu
b43vvdr+state:results

but I don't understand. He says it was because our Java wrappers weren't 
unique. They were being recycled?

on a moving gc the identity hash is not unique, id() needs to be unique.
So not using it is a matter of correctness not speed.

But this issue is too terse for me to be sure what is talking about.

That's the answer I was looking for, thanks Samuele