Revision history for Perl extension POE::Component::IKC.

0.2303 16 Jan 2013
    - Cleaned up cookie crumbs
    - Removed erroneous call to _select_define in Server::accept
        Thank you rbhardwaj (rt82685)

0.2302 26 Aug 2011
    - Forgot IKC::Protocol  GAH!
        Thank you Gregor Herrmann

0.2301 26 Aug 2011
    - Bump copyright year

0.2300 26 Aug 2011
    - Fixed thunk checking for 1.311
    - Added IKC::Timing, temporarily
    - Created a new, faster protocol (IKC0) which sends all info on one line
      as opposed to multiple lines as before (IKC)
    - Channel session will detach itself from Client session

0.2200 08 May 2009
    - Better concurrency handling
    - We now have a reusable thunk session
    - Added test cases for above
    - Reworked t/30_concurrency.t Seems it was failing on Win32
    - Added ClientLite->responded()

0.2102 06 May 2009
    - IKC::Channel uses presence of IKC alias, not {shutdown}, to unregister
    - Added a _child to shut up POE::Session
    - Changed the DEPRECATED messages that fluster everybody into
        pointers to IKC/monitor.  I'm not going to remove on_connect nor
        on_error.
    - Allow IKC::Server port=0

0.2101 01 May 2009
    - IKC::Channel stops work during shutdown
            Fixes http://rt.cpan.org/Public/Bug/Display.html?id=44104
            Thank you, acferen [...] yahoo.com

0.2100 01 May 2009
    - LICENSE in Makefile.PL
    - Added concurrency to IKC::Server
    - Moved all tests to Test::More 
    - Added POD tests
    - Updated copyright year
    - Cleanup dist
    - Removed checks for ancient versions of POE
    - Require POE 1.000

0.2003 03 April 2009
    - IKC::Channel->spawn returns session ID
    - Only keep channel IDs in IKC::Responder

    - Several changes to improve shutdown behaviour :
        - Register channels with the IKC::Responder before they negociate
        - Shutdown a IKC::Client when its Channel closes, so that IKC 
            shutdown will also shutdown the channels
        - Give Clients aliases to help debuging
        - Shutdown IKC::Server when the last channel goes away and 
            there's no wheel
    
0.2002 26 November 2008
    - Added call to $kernel->_data_sig_initialize, otherwise child processes
        will never exit

0.2001 16 January 2008
    - Added copyright and license notices to all POD, as requested by
        Ernesto Hernandez-Novich.

0.2000 29 November 2007
    - FreezeThaw::thaw has started to return an array.  We only want the first
        element... 
            Thank you Faiz Kazi
    - Better timeout handling in IKC::ClientLite.
    - Documented the above

0.1905 8 August 2007
    - It was possible for a IKC::Server child process to inherit
        some delay()s, clear those also.
    - Clear all delay()s at the least provocation (INT, TERM, errors, etc)
    - Created IKC::Server::_delete_wheel
    - Use POE::Component::Daemon->peek

0.1904 16 November 2006
    - Channel now reacts better to IKC1 connections

0.1903 16 November 2006
    - Added t/01_specifier.t
    - Fixed callback example in IKC.pod
    - Allow 'poe:'-less specifiers

0.1902 1 November 2006
    - Forgot $VERSION in IKC.pm

0.1901 3 October 2006
    - Tweak the POD
    - Forgot IKC.pm in MANIFEST

0.1900 3 October 2006
    - The memory leaks (big and small) are now verified as gone with
        POE 0.37 and perl 5.8.8.  (Thank you dw for test case.)
    - Remove Channel aliases.  DUH!

0.1804 29 August 2006
    - Added IKC.pm
    - Check the POE leak

0.1803 2 November 2005
    - Use Scalar::Util
    - Better debug messages

0.18 21 June 2005
    - ClientLite was looking for freeze before nfreeze, which broke 
        cross-platform usage (Thank you Philip Dobranski)
    - Channels weren't being shut down properly after X connections
    - Channels couldn't be shut down from a Client

0.16 - 0.17 Internal releases

0.1501 23 Mar 2005
    - Fixup some of the DEBUG messages
    - Changed from use of internal {pending} to wheel->get_driver_out_octets
    - Added WORK AROUND for wheel leak issue
    - LocalKernel now removes it's alias, to be tidy
    - Removed myself element of Responder's heap.  What was it going to be
        used for?
    - Responder's {monitors} wasn't being cleaned up, fixed
    - Server _select_define now uses {pause,resume}_select, which
        is better behaved then the internal hackery previous
    - Slight reworking of pre-forking code
    - All tests pass

0.15 26 May 2004
    - Added on_error, which is automatically deprecated for the monitor stuff
    - Fixed some of the test
     13 May 2004
    - Moved to signal_handled

0.14 17 Oct 2002
    - Nothing changed

0.14pre3 18 Oct 2002
    - Uses sig_handled() where appropriate

0.14pre2 16 Oct 2002
    - Now uses *Event rather then *State

0.14pre1 2 may 2002
    - inform_monitor is now an event
    - IKC::Proxy post()s to inform_monitor in _start
    - IKC::Proxy call()s inform_monitor in _stop
    - Merged up alias listing in publish/retract

0.1303 26 Oct 2001
    - IKC::Responder->spawn returns true on success
    - IKC::Responder->spawn documented
    - Checking to make sure a publishing session has an alias.  if not, use ID
    - First babysitting happens at startup, doesn't wait
    - check_kernel brought in line w/ new kernel internals

0.1302 6 Sept 2001
    - Meaningless version change

0.1301 24 July 2001
    - unix domain paths in kernel names now have / and \ turned into _
    - Local kernel gets many names
    - Make sure sockets are open before syswrite during neg phase of 
        ClientLite, allowing better error handling
    - IKC::Responder::spawn now checks that only one responder exists
    - alias_list is only supported after 0.15
    - Added win32 support (WSAEAFNOSUPPORT)
    - IKC::Responder has better state dumping
    - Better specifier matching
    - Added test for ClientLite

0.13  13 July 2001
    - Added and repaired serializer selection for clients
    - Added doco for ClientLite
    - Sungo is happy with this release... onto CPAN it goes.

0.13pre9  5 July 2001
    - Doco fixes

0.13pre8  4 July 2001
    - Fixed the deprecated use of defined(@array) in IKC/Responder.pm
    - Added doco about SENDER
    - create_ikc_channel now takes an arrayref of aliases, 
      so that IKC::Client and IKC::Server can accept a longer list of
      pre-defined names.
    - IKC/post2 now expects SENDER to come before [state, args]
    - Added doco to the effect that callback states are temporarily published
    - Updated code that turned off the socket_factory... Dngor had
      changed SocketFactory's implementation on me!  This is what you get
      for doing bad stuff.

0.13pre7  20 June 2001
    - Added IKC/published, for sungo
    - Fixed doco a bit
    - Added a prereq for POE 0.1403

0.13pre6  6 June 2001
    - ADDED TESTS!
        Woot woot!  Now make test verifies something meaningful.
    - No longer thunking sessions that don't publish any states
    - Finally did unsubscribe... turns out retract was available all along
    - Added IKC::{Client,Proxy,Responder,Server}->spawn method, to 
        be used in favour of create_ikc_*mumble*
    - Added IKC/post2 .... internal use only
    - Added FlushedState handler to IKC::Channel.  This way, wheel isn't
        destroyed until all data is flushed.  This means you can post
        events to remote kernels and then IKC/shutdown right away.  I hope.

0.13pre5  4 June 2001
    - Added IKC/shutdown and attendant bouffonery (see shut-{server,client})
    - IKC::Proxy for 'poe://somekernel/somesession' is now also aliased
      as 'somesession'
    - Added unix-domain sockets... don't know if i've got this right
        see u{server,client}

0.13pre2-4  May 2001
    - Fixed doco and monitor stuff

0.13pre1 19 Apr 2001
    - Added the IKC/monitor stuff
    - Added some doco
    - ikc-use.txt is now IKC.pod

0.12  March 2001
    - Updated Changes, README.  Added FUTUR
    - Improved warning messages
    - IKC::ClientLite::disconnect no longer calls IKC/unregister to remote
      (doing it was unneed, server does it automatically when socket dies)

0.11  Nov 16 2000
    - IKC/ClientLite now sets $/ and $\ to "\r\n"
    - wantarray() now gets propagated... is this tested?
    - babysitting now knows of "defunct" children
    - children now turn off SIGCHLD and SIGINT... dunce :)
    - check_kernel debuging subroutine

0.10  May 26 2000
    - Added $$ to all the warnings to help isolate problems
    - Converted IKC::Server to package methods
    - Split signals into sig_INT, sig_CHLD...
    - Improved babysitting code (rogues)

0.09  24 May 2000
    - Fixed things that POE 0.0906 broke
    - Added more babysitting code to pre-forking server.  Will document soon.
    
0.08  24 Jan 2000
    - Fixed non-forking server code :)
    - 0.0808 changed Wheel::SocketFactory internals

0.07  Dec 1999
    - Added pre-forking server
    - Added IKC::ClientLite::post_respond

0.06  Saturday Dec 18 1999
    - Fixed some warning conditions

0.05  Saturday Nov 06 1999
    - Chanels are now being closed when a remote kernel unregisters
      (this was a TODO, but previously only EPIPE or socket errors
      would cause this to happen and it was leaking ram)
    - Small cosmetic changes

0.04  Sep 23 1999
    - IKC::ClientLite
    - Some changes :)

0.03  Thr May 20 22:32 1999
    - Added the object interface
    - Documented the publish/subscribe
    - Added access restrictions so that only published
      states may be posted

0.02  Wed May 19 xxxx 1999
    - Added publish/subscribe 

0.01  Tue May 11 21:18:04 1999
	- original version; created by h2xs 1.18