Hi Peter,
Yes, I had considered a mutex pool. One problem is that boost::shared_ptr is header-only, and must remain so for backward compatibility reasons.
Well, that would preclude such an optimization, though I am not sure how adding a .cpp file would break code (just makefiles<g>).
Another problem is that a CRITICAL_SECTION needs to be dynamically initialized before use, but the initialization itself also needs synchronization.
Unless it happens at global construction time as with Singleton. You are correct that the pseudo code I posted would not work if mutex == CRITICAL_SECTION.
It is easy to apply this optimization to std::tr1::shared_ptr, though, since standard libraries don't need to be header-only and can initialize things before everything else. ;-)
Good news indeed! Best, Don ===== __________________________________ Do you Yahoo!? Read only the mail you want - Yahoo! Mail SpamGuard. http://promotions.yahoo.com/new_mail