Anything involving multiple threads segfaults, such as Dispatchers and Queue
hooks.  It may work with various combinations of perl version and perl
thread model (untested).


Fault Tolerance unimplemented


Distributed Queuing unimplemented


Virtual Circuits unimplemented


Secure Daemon connections unimplemented


Cm::Transport connectToRelayAgent/disconnectFromRelayAgent return INVALID_ARG
instead of ARG_CONFLICT as specified in the TIB/Rendezvous documentation.
(It's just a documentation error.  I should really email them about that ...)


IO should probably take socketId => \*GLOB instead of a fileno


Queue hook should maybe be set up like Event's onEvent callback?


Cm::Transport reviewLedger callback should maybe be set up like Event's
onEvent callback?  Perhaps only if $callback parameter is undef?


Gotta do something to abstract out createInbox ...


Tibco::Rv::Msg::DateTime
doesn't seem to hold its nsec value ...  I think it's a bug w/tibrv


Tibco::Rv::Dispatcher::DESTROY
If idleTimeout is something other than WAIT_FOREVER or NO_WAIT, and that
idleTimeout is reached, then tibrvDispatcher_Destroy is called without
calling the Perl DESTROY method, putting the dispatcher in a bad state.  No
fix is contemplated.  So don't access a Dispatcher after it times out.
I'm also thinking that DESTROY will throw an exception when the Perl object
gets garbage collected in this situation.  But that's not really a big deal.


Tibco::Rv::Event::DESTROY
tibrvEvent_DestroyEx has a callback, but no closure data (unlike
tibrvQueue_DestroyEx -- I wonder why?).  Since Perl callbacks are stored
internally in the closure data, Tibco::Rv does not support the
tibrvEvent_DestroyEx callback.  It could be hacked in by storing the DESTROY
callback along with the onEvent callback in a hash or something, and passing
both callbacks into the Perl constructor.


tibrvMsg_SetHandlers
Not supported.  I'm sure no one wants this stuff, anyway.