CAS is broken

[Tomash Brechko]

On Sun, Nov 11, 2007 at 23:35:55 -0800, Dustin Sallings wrote:
> 	An object isn't reused during a write.  A new item is allocated when 
> you begin a set.  The old one is served out until the new write is  
> finished (at which point the old one is added back to the free pool).
> 
> 	I wasn't as worried about the thread synchronization because
> 
> 	a) The abstraction makes it trivial to add a mutex if necessary.
> 	b) It's likely already in a mutex (although I didn't check)

Yes, I already understood my mistake.  CAS counter should be global,
and no fast path is possible in the client as I thought at first.


-- 
   Tomash Brechko