2004-11-09  Al Riddoch  <alriddoch@zepler.org>

	* Release 1.2.1. Interface version 2.0.0, or 3.0.0 with sigc++ 2.0.

2004-11-07  Al Riddoch  <alriddoch@zepler.org>

	* test/testConnection.cpp: Add a sigc++ header required to build
	  this test against sigc++ 2.0.

	* Update configure.ac to detect (and prefer) sigc++ 2.0.

2004-11-07  Al Riddoch  <alriddoch@zepler.org>

	* Eris/SignalDispatcher.h: Add a sigc++ header required to build
	  against sigc++ 2.0.

2004-06-06  Al Riddoch  <alriddoch@zepler.org>

	* Fix devel dependencies in rpm spec.

2004-06-03  James Turner  <james@worldforge.org>

	* XCode project updates

2004-05-19  Al Riddoch  <alriddoch@zepler.org>

	* Release 1.2.0, interface version 2.0.0

2004-05-19  Al Riddoch  <alriddoch@zepler.org>

	* Clean up pointless use of a few macros in configure.ac. Trim
	  version down to 3 numbers, rather than 4.

	* Fix deps in rpm spec, and make sure to use the right compiler
	  flags.

2004-05-19  James Turner  <james@worldforge.org>

	* Version number tweaks in anticipation of a release
	  in the next day or two
	* Fix eris.dox.in as suggested by Michael Koch
	* Add some cvsignore entries

2004-05-18  Al Riddoch  <alriddoch@zepler.org>

	* Tweak getting of canonical system.

2004-05-15  James Turner  <james@worldforge.org>

	* Eris/World.cpp, Eris/World.h, Eris/SignalDispatcher.h: Fix how 
          SIGHT(CREATE) is handled, so the value we emit via GotTime
          is correct. This entailed adding a SignalDispatcher3<> template.

2004-05-13  James Turner  <james@worldforge.org>

	* Eris/World.h, Eris/World.cpp: Add a GotTime signal, and emit it
          in response to sight(entity), sight(move) and sight(set). This
          entailed using SignalDispatcher2 in a few more places.

2004-05-04  Al Riddoch  <alriddoch@zepler.org>

	* Eris/TypeDispatcher.h, Eris/Dispatcher.cpp: Add new dispatcher
	  that deals with objtype being "object" or "obj".

	* Eris/Connection.cpp, Eris/Lobby.cpp, Eris/World.cpp: Use the
	  new dispatcher so we can talk to Atlas-C++ 0.6.

2004-04-19  Al Riddoch  <alriddoch@zepler.org>

	* Eris/PollDefault.cpp: Ensure that flags don't get broken
	  if exception gets thrown in PollDefault::poll by catching
	  and rethrowing.

2004-03-06  Al Riddoch  <alriddoch@zepler.org>

	* Eris/Entity.cpp, Eris/Entity.h: Make getProperty() member function
	  const.

2004-03-04  Al Riddoch  <alriddoch@zepler.org>

	* Bump the interface version so runtime libs for transitional
	  don't interfere with stable.

	* Eris/Metaserver.cpp, Eris/Metaserver.h: Add compatability calls
	  with development Eris so we can push ahead with UI features
	  to test development Eris.

Thursday, 5th February 2004, Al Riddoch <alriddoch@zepler.org>

	* Update dependencies and header list in the rpm spec.

	* Add a few EXTRA_DIST files, include spec, and pgxproj files.

	* Release 1.1.1, interface version 1.0.0

Thursday, 5th February 2004, James Turner

    * Fix duplicate characters in the Player's list, by discarding
      duplicate SIGHTs and locking out calls to refreshCharacterInfo while
      a refresh is already in progress
    * Switch the internal character storage to be a map, keyed by entity ID,
      and make the accessor method (Player::getCharacters()) return a
      const reference.
    * Various other whitespace changes in Player.cpp
    * Bump version to 1.1.1 in anticipation of release
    * Set required WFMath version to 0.3.3 for coordinate conversion helpers
    
Saturday, 24th January 2004, Al Riddoch <alriddoch@zepler.org>

	* Eris/UIFactory.h, test/testUtils.cpp: Move cassert includes
	  around a bit, for fun and profit.

Thursday, 22nd January 2004, Al Riddoch <alriddoch@zepler.org>

	* Eris/Exceptions.h Eris/Exceptions.cpp: New files to contain
	  exception class declarations and definitions.

	* Eris/Types.h: Remove exceptions, and clean up types.

	* Re-organise includes to use new exceptions header.

Thursday, 22th January, Ron Steinke <rsteinke@w-link.net>

	* Remove ::Instantiate() calls from test/*

Thursday, 22nd January 2004, James Turner <james@worldforge.org>

    * Remove use of deprecated ::Instantiate() calls
    * Require Atlas-C++ 0.4.93 (latest transitional)

Thursday, 22nd January 2004, Al Riddoch <alriddoch@zepler.org>

	* Eris/typeService.cpp: Update type service to use Atlas-C++
	  Class() method, as default constructor now creates an instance.

Monday, 19th January 2004, Al Riddoch <alriddoch@zepler.org>

	* Many more include fixups, mostly relating to sigc++.

Sunday, 18th January, Ron Steinke <rsteinke@w-link.net> (don't you love time zones?)

	* test/testPlayer.cpp: Added an #include so the tests build again.

Monday, 19th January 2004, Al Riddoch <alriddoch@zepler.org>

	* Reorder includes, and elimate unnecessary includes by using
	  forward declarations throughout.

Sunday, 18th January, Ron Steinke <rsteinke@w-link.net>

	* Eris/Avatar.*: Renamed the move() functions to
	  moveToPoint() and moveInDirection() to avoid
	  the Point/Vector argument overload.

Sunday, 18th January, James Turner <james@worldforge.org>

	* Patch from Ron Steinke : make Factory.h use forward declared Atlas
	  types, to avoid dragging in huge chunks of Atlas::Objects and all the
	  Atlas::Mesage stuff.

Saturday, 17th January, Ron Steinke <rsteinke@w-link.net>

	* Eris/Avatar.*: Added velocity-based move() methods.

Sunday, 11th January, Ron Steinke <rsteinke@w-link.net>

	* Eris/Metaserver.*: On second thought, just move
	  the snprintf stuff into Metaserver.cpp. That way
	  Eris should continue to build fine, but dependent
	  software won't get hit with it.

Sunday, 11th January, Ron Steinke <rsteinke@w-link.net>

	* Eris/Metaserver.h: Put an #ifndef __MINGW32__ around
	  the snprintf and vsnprintf #defines.

Tuesday, 6th January 2004, Al Riddoch <alriddoch@zepler.org>

	* Eris/World.h, Eris/World.cpp: Add a signal to indicate if
	  if a player has successfully taken control of a new or existing
	  avatar.

Saturday, 20th December, Ron Steinke <rsteinke@w-link.net>

	* Don't need Avatar::GotEntity, World::Entered does
	  the same thing.

Saturday, 20th December, Ron Steinke <rsteinke@w-link.net>

	* Added a GotEntity signal to Avatar, so we can connect
	  callbacks to the character's entity when it arrives.

Friday, 19th December, Al Riddoch <alriddoch@zepler.org>

	* Remove Serial from rpm spec as it is not required, and messes up
	  deps.

Thursday, 19th December, James Turner <james@worldforge.org>

    * Remove ancient socket references from Eris/Types.h - long
      since deprecated in favour of skstream

Monday, 8th December, James Turner, <james@worldforge.org>

    * Fix a warning on Mingw32, due to me using -1 in BaseConnection,
      when I should have used the constant supplied by skstream.

Friday, 5th December, Al Riddoch <alriddoch@zepler.org>

	* Fix LDADD for testEris so it works right.

Friday, 5th December, Al Riddoch <alriddoch@zepler.org>

	* Fix test/Makefile.am so the tests build but don't run.

Tuesday, 2nd December, Al Riddoch <alriddoch@zepler.org>

	* Supress unknown pragme warnings, and add canonical macros.

Thursday, 27th November, James Turner <james@worldforge.org>

    * Apply Ron's patch for transforming WFMath objects
      based on an entity's pos and orientation.

Monday, 24th November, James Turner <james@worldforge.org>
    * Add accessor for Person objects directly on Room.
    
Monday, 24th November, Al Riddoch <alriddoch@zepler.org>

	* Eris/Avatar.cpp: Fix all uses of send() to pass a reference
	  rather than a pointer. Why did this compile?

Sunday, 23nd November, Ron Steinke <rsteinke@w-link.net>

	* Added take(), touch(), say(), move(), and place()
	  methods to Avatar, to complement the existing drop()
	  method.

Sunday, 2nd November, Ron Steinke <rsteinke@w-link.net>

	* Replaced Time::Stamp class with a typedef to
	  WFMath::TimeStamp (which was originally a copy
	  of Time::Stamp).

Sunday, 26th October, Ron Steinke <rsteinke@w-link.net>

	* Added the capacity to have multiple timeouts with
	  the same name, as long as they are owned by different
	  object instances. This fixes a problem with the code
	  that allows multiple simultaneous Connection instances.

Thursday, 23rd October, Simon Goodall <simon@simongoodall.co.uk>
	* Fixed configure.ac to use eris-poll-glib-1.2.pc in second glib test

Monday, 20th October, James Turner <james@worldforge.org>

	* Add ProjectBuilder project - currently requires a modified
	  libsigc++ framework which I'll upload to FTP at some point.

Saturday, 20th September, Ron Steinke <rsteinke@w-link.net>

	* Fixed test code so it acutally builds, mostly Atlas
	  stuff and changing pollDefault to PollDefault (which is
	  odd, because it's _never_ been pollDefault).

Saturday, 23rd August, Al Riddoch <alriddoch@zepler.org>

	* Eris/BaseConnection.cpp: Switch to using skstream::isReady()
	  instead of the deprecated is_ready().

Friday, 22th August, Ron Steinke <rsteinke@w-link.net>

	* Made Poll::new_timeout_ static to keep the Timeout
	  constructor from instantiating the poll. Otherwise
	  static timeouts instantiate PollDefault before
	  we can set the poll instance.

Friday, 22nd August, James Turner <james@worldforge.org>

	* Rename configure.in to configure.ac, since we require the newer
	  version. This stops some autoconf versions moaning about the name,
	  and helps scripts which detect which autoconf to run by
	  examining configure.in
	* Also add an AC_PREREQ, just for paranoia's sake.
	* Fix autogen.sh for the new name

Sunday, 17th August, Ron Steinke <rsteinke@w-link.net>

	* Eris/UIFactory.*: Fixes to UI code for change
	  to atlas-transitional.

Monday, 11th August, James Turner <james@worldforge.org>

	* Fix a dumb bug I introduced in June, in handling server-
	  commaned LOGOUTs. Essentially, I failed to pickup the
	  anonymous ClassDispatcher below :op:, and hence was
	  dispatching Player::recvRemoteLogout for every op
	  recived. Solution is too create an anonymous ClassDispatcher
	  ourselves, pending a re-think of anonymous dispatchers
	  at some point in the future.
	  
	  Thanks to Al Riddoch for the catch.

Monday, 11th August, Al Riddoch <alriddoch@zepler.org>

	* Eris/PollDefault.cpp, Eris/PollDefault.h: Fix more incorrect mangling
	  of Poll by conversion script.

Saturday, 9th August, Al Riddoch <alriddoch@zepler.org>

	* Switch all the build files over to using 1.2 API suffix.

	* bindings/polls/glib/PollGlib.h, bindings/polls/glib/PollGlibFD.h,
	  bindings/polls/glib/PollGlibSource.h: Fixed some incorrect
	  mangling of Poll by the Atlas-C++ conversion script.

Saturday, 9th August, James Turner <james@worldforge.org>

	* Convert to Atlas-C++ transitional API. Do not use unless
	  you know what you're doing. Arrr!
	* Also convert to skstream-0.3 and require wfmath 0.3

Tuesday, 22nd July 2003, Ron Steinke <rsteinke@w-link.net>

	* Added UI Factory code, based on the new experimental
	  Janus UI Atlas entities. You need the experimental
	  janus branch of Atlas-C++ to compile it, so it's
	  wrapped in a nice-freindly ./configure check.

Monday, 23rd June 2003, Al Riddoch <alriddoch@zepler.org>

	* Fix Makefile.ams so that distcheck works.

Friday 13th June, James Turner <james@worldforge.org>

	* Fix compilation of the typeService stuff, this once again
	  provides the danger in committing while in a hurry to play
	  games ;-)
	* Change the dependancy maps to key off of a TypeInfoPtr instead
	  of the string name.
	* Moved all the remaining TypeService methods out of TypeInfo.cpp
	  and into typeService.cpp where they belong

Monday, 9th June, James Turner <james@worldforge.org>

	* Fix a waring in Entity.h spotted by rsteinke
	* Rename TypeInfoEngine to TypeService, move
	  to seperate files, and re-factor TypeInfo code
	  to know much less about the TypeService's
	  internal state (which is now private)
	* Rename various Type system methods as suggested
	  by Al Riddoch to be more descriptive and useful

Friday, 30th May 2003, Ron Steinke <rsteinke@w-link.net>

	* Fixed a missed Interface -> UserInterface change
	  in Player.h.
	* Added a <cassert> #include in atlas_utils.h, so
	  the server test builds under g++ 3.3.

Sunday, 25th May 2004, James Turner <james@worldforge.org>

	* Start to migrate the test code away from CppUnit,
	  since ithe typeinfo magic it uses upsets GCC < 3.2
	  Instead, switch to some fairly basic internal
	  helpers, since we never did much with CppUnit
	  anyway.
	* Start adding TypeInfo support to the stub server
	  (using hard coded data, no dependancy on an
	  external source), so it be a bit more realistic.
	* Rewrote testPlayer to test more things, not use CppUnit,
	  and generally be smarter. Oh, and it uses StubServer,
	  since StubConnection is a bad idea and will be
	  going away shortly.

Tuesday, 13th May 2003, Ron Steinke <rsteinke@w-link.net>

	* More perl bindings memory management stuff,
	  primarily handling the World::Destroyed signal.
	* Used templates to simplify the perl typemap
	  stuff a bit, changed the #includes in the modules
	  to only include the type conversion code
	  each needs.

Tuesday, 13th May 2003, Ron Steinke <rsteinke@w-link.net>

	* Gave the perl bindings better memory management.
	  The biggest change is that rooms are now passed
	  around as a lobby pointer and id string, so we
	  don't get seg faults if room references go stale.
	  Most of the rest is incrementing/decrementing
	  the refcounts of the perl SV's that own Connection
	  and Player instances.
	* Added some getConnection()/getLobby()/getPlayer()
	  accessors to various classes, as needed for the
	  above memory management code.
	* Added an Interface class, for the beginning
	  of incorporating handling of the soon-to-be-defined
	  Atlas interface ops (old janus functionality).

Saturday, 10th May 2003, Al Riddoch <alriddoch@zepler.org>

	* Incremented version to 0.9.8.1 for a patch release with
	  working perl bindings.

Saturday, 10th May 2003, Ron Steinke <rsteinke@w-link.net>

	* Changed perl bindings build stuff to account
	  for the @WFMATH_CFLAGS@ @ATLAS_CFLAGS@ -> @WF_CFLAGS@
	  change
	* Added a check for the SigC module to the check for
	  the sigcperl lib in configure.in
	* Got rid of OLD_SIGC_1_0 stuff in bindings, since we're
	  now requiring libsigc++ 1.2

Friday, 9th May 2003, James Turner <james@worldforge.org>

	* Incremented version to 0.9.8 for stable release.

Thursday, 8th May 2003, Ron Steinke

	* Changed Al's mkdir -p fix in bindings/Makefile.am
	  to test -d || mkdir, since that's the way autoconf
	  does it and it's sure to be portable.

Thursday, 8th May 2003, James Turner <james@worldforge.org>
	
	* Don't make or install eris-config, eris.m4 or 
	  eris-config.1 (the man page). The actual files are still
	  there, will be removed fairly sooon.

Thursday, 8th May 2003, Al Riddoch <alriddoch@zepler.org>

	* Use $(top_buildir) to refer to the top of the build structure,
	  and remove reference to old src dir from test Makefile.

Thursday, 8th May 2003, Al Riddoch <alriddoch@zepler.org>

	* Re-order configure.in so the version variable works.

	* Sort out perl bindings dist-hook rule so it works if the
	  perl directory already exists.

Thursday, 8th May 2003, Al Riddoch <alriddoch@zepler.org>

	* test/tests.cpp: Skip the incomplete Player and Lobby tests.

Thursday, 8th May 2003, Al Riddoch <alriddoch@zepler.org>

	* test/tests.cpp: Return with the correct exit status if the
	  tests fail.

Thursday, 8th May 2003, James Turner <james@worldforge.org>

	* Updated configure.in to be Autoconf 2.5 happy, removed lots
	  of deprecated macros
	* We now rely on pkg-config versions of sigc++ and all the 
	  WorldForge libs, so configure is smaller and all the
	  AM_PATH_xxx copies in acinclude.m4 go away and die.
	* Made the WFmath handling sane too (we had extra -lwfmath
	  and such in ERIS_LIBS, which is unecessary since pkg-config
	  will pick that stuff up for us)

Thursday, 8th May 2003, Ron Steinke

	* Added typemaps to perl MANIFEST
	* Added a dependency on the build library to
	  bindings/perl-makefile.stamp, so the modules
	  get rebuilt if the lib does

Thursday, 8th May 2003, Ron Steinke

	* Missed bindings/perl/compile_flags.in. Note that
	  make distcheck will still not work when building
	  the perl bindings, since makemaker can't handle
	  having $(srcdir) and $(bindir) be different

Thursday, 8th May 2003, Ron Steinke

	* Added a dist-hook so the perl bindings make it into
	  the automake tarball
	* Fixed the way the perl bindings include the eris
	  headers to match the new way of doing things

Wednesday, 7th May 2003, James Turner <james@worldforge.org>

    * Updated README, INSTALL and AUTHORS in anticipation of
      a stable release fairly soon.

Tuesday, 6th May 2003, Ron Steinke

	* Further perl bindings build fixes

Tuesday, 6th May 2003, Ron Steinke

	* Fixed a problem with the perl bindings not getting
	  the atlas compile flags, generally cleaned up
	  the passing of those flags from autoconf to
	  the perl makemaker build system

Friday, 2nd May 2003, Al Riddoch

	* Move code from src subdir to Eris.

	* Add second .pc file for glib headers, sort out which version of glib
	  it depends on, and sort out to handle it in the rpm.

Tuesday, 22nd April 2003, James Turner

	* updated README / INSTALL to be up-to-date for the next
	  release, and converted to plain text (from Abiword)

Monday, 24th March 2003, Al Riddoch

	* src/Dispatcher.cpp: Handle missing TO on operations.

Monday, 24th March 2003, Al Riddoch

	* bindings/polls/glib/Makefile.am: Move bindings headers to the new
	  pkgconfig compliant header directory.

Saturday, 1st Febuary 2003, Al Riddoch

	* Convert to pkgconfig.

Friday, 31st January 2003, Al Riddoch

	* Fix Atlas-C++ compiled flags.

Thursday, 30th January 2003, Al Riddoch

	* src/Lobby.h: Make destructor virtual.

	* src/Room.cpp, src/Room.h: Simplify to one constructor.

	* src/World.cpp: Ensure talk dispatcher names are unique.

Monday, 16th December 2002, Ron Steinke

	* src/Avatar.h, src/BaseConnection.h, src/ClassDispatcher.h,
	  src/Entity.h, src/Metaserver.h, src/Player.h, src/Property.h,
	  src/Room.h, src/SignalDispatcher.h, src/Timeout.h,
	  src/TypeInfo.h, src/Wait.h, src/World.h: Changed classes
	  inheriting 'public SigC::Object' to inherit 'virtual public
	  SigC::Object' instead, so multiple inheritance will work

Wednesday, 14th December 2002, Al Riddoch

	* src/Dispatcher.cpp: Take reference to StdBranchDispatcher before
	  calling its subdispatchers so it doesn't get deleted before its
	  finished. Fix use of iterators.

	* src/Dispatcher.h, src/Connection.cpp: Make reference handling
	  methods public, and make connection hold a reference to the
	  root dispatcher so it doesn't get deleted.

Wednesday, 12th December 2002, Al Riddoch

	* src/Avatar.cpp: Fix initialisation of val.

Wednesday, 11th December 2002, Al Riddoch

	* test/testPlayer.cpp: std:: namespace fixes.

	* configure.in: Handle interface version correctly.

Monday, 2nd December 2002, Ron Steinke

	* Fixed check for Perl bindings so it can detect SigCPerl >= 0.2.0,
	  which uses pkg-config
	* Fixed errors in Perl binding build, made some changes necessary
	  to accomodate libsigc++-1.2

Friday, 29th November 2002, Ron Steinke

	* Added PollGlib, with code taken from silence. This is the poll
	  for use with gtk+, it's called PollGlib because it only depends
	  on the lower level glib library.

Tuesday, 27th November 2002, Ron Steinke

	* Integrated Perl bindings into the build with --enable-perl.
	  Rebuild dependencies still don't work properly, and make dist
	  isn't integrated between the automake and makemaker parts.

Tuesday, 26th November 2002, Ron Steinke

	* Added a missing forward declaration of World to Factory.h

Monday, 18th November 2002, Ron Steinke

	* Changed a bunch of accessors that returned std::string
	  to return const std::string& instead

Thursday, 14th November 2002, Ron Steinke

	* Changed the second argument of Meta::ReceivedServerInfo
	  from ServerInfo to const ServerInfo&, fixed a typo

Tuesday, 12th November 2002, Ron Steinke

	* Added an optional offset to Avatar::drop().
	* Split Avatar::InvChanged into Avatar::InvAdded
	  and Avatar::InvRemoved.

Wednesday, 6th November 2002, Ron Steinke

	* Added a signal handler to set Avatar's _entity member
	* Added some inventory handling capabilities to Avatar

Tuesday, 5th November 2002, Ron Steinke

	* Got Avatar working properly with OpRefnoDispatcher,
	  and changed existing references in the code to the
	  op:ig:foo dispatcher to use the appropriate per-World
	  dispatcher id. Taking/creating characters works again.

Sunday, 3rd November 2002, Al Riddoch

	* configure.in: Clean up, and disable static by default.

Sunday, 3rd November 2002, Al Riddoch

	* Add rpm spec.

Friday, 1st November 2002, Al Riddoch

	* src/World.cpp: Fix calls to SigC::slot so it is sigc++ 1.2
	  compatable.

Friday, 1st November 2002, Ron Steinke

	* Change ArgumentDispatcher to (new) OpRefnoDispatcher in Avatar
	* Change std::cerr to log(LOG_DEBUG,...) for debugging code

Friday, 1st November 2002, Al Riddoch

	* src/World.cpp: ISO C++ does not allow default value for arg to be
	  specified in function definition.

Wednesday, 30th October 2002, James Turner

	* Commit Ron's Avatar patches : currently got a bug
	  with the IG-transition INFO op, so don't update unless
	  you're doing dev work. Also bumped the version to 0.9.7
	  since this is an API change (rsteinke has diffs for silence
	  and sear, very minor).

Monday, 28th October 2002, Al Riddoch
	* Fix dispatcher removal for the imaginary emote ops fix made on the
	  9th. October.

Thursday, 24th October 2002, Al Riddoch
	* Add checks so it is possible for a connection to send and
	  receive when its status is DISCONNECTING.

Thursday, 24th October 2002, Simon Goodall
	* Clean up DebugDispatchers when deleting Connection object

Wednesday, 09th October 2002, Al Riddoch

	* src/ArgumentDispatcher.cpp: Prevent ArgumentDispatcher from
	  aborting if the object it is passed has no args.

	* src/Room.h, src/Room.cpp: Rework handling of imaginary emote
	  ops to be a bit more general.

Tuesday, September 17, Ron Steinke

	* Backed out fix for Connection seg fault, which was
	  causing problems

Friday, September 13, Ron Steinke

	* Added a check to Player::netConnected() that the current
          action is null before calling internalLogin(). This
          fixes a double login problem if someone else calls
          Player::Login() in a BaseConnection::Connected
          signal handler before the signal gets to
          Player::netConnected().

Sunday, September 1, Ron Steinke

	* Enhanced locking in Connection to prevent a segfault when
	  Connection::disconnect() is called inside a signal emitted
	  by BaseConnection::recv() (e.g. Player::LogoutComplete)
	* Added a lock in PollDefault::poll() to prevent reentrancy,
	  so we catch it sooner instead of hitting an assert()
	  in Dispatcher

Wednesday, August 28th, James Turner

	* Fix a bug (found by Simon) in Connection::gotData, introduced
	  by switching the lifetime of the underlying tcp_socket_stream. Basically
	  we weren't checking if the _stream existing prior to calling isReady.
	  If the meta-server is being used, and a connection object is created but
	  not connected, then this bug (a segfault occurs), since the Connection
	  is registered with the Poll instance and network activity can happen.
	  
	* Change the meta-server code to use the same life-time rationale for
	  it's UDP stream that BaseConnection uses for the Atlas stream. Basically
	  create the stream on refresh() and not before. This changes the default
	  behaviour : creating a Meta instance will not trigger a list query. This allows
	  the client to attach to the relevant signals before any failures can
	  occurr (thanks Simon)
	  
	* Fixed game server list persistance (I hope). If a failure or cancel
	  occurs while refreshing the list, Eris will revert back to the last
	  'good' list it recieved (if one exists).
	  
	* Converted dispatcher clean up in Lobby dtor to use removeIfDispatcher instead
	  of lame try { .. } catch {}  blocks.
	  
	* Added an explicit greater-than comparisom to Timestamp, so the tests compile.
	  Did GCC 2.9.x implicitly use !(a > b) before? Certainly 3.1 seems not too.

Thursday, August 8th, Al Riddoch

	* Fixed sigc++ use so it complies with published APIs, and put
	  conditionals in to pick up headers for sigc++ 1.0 or 1.2.

Saturday, August 3rd, James Turner

	* Change BaseConnection to create a new tcp_socket_stream
	  for every connect. (And, obviously, hardDisconnect deletes it)
	  This should fix handling of stream errors, since there is no
	  safe way to clear the error code apparently.
	  This change is low impact code-wise (only changes in
	  BaseConnection.cpp), but needs good testing.

Friday, August 2, Michael Koch

	* Patch to declare operators in header file and made method
	  headers equal to their declaration in header file.

Monday, July 29, Ron Steinke

	* Patch to allow multiple Connection instances. Lobby and
	  TypeInfoEngine are per Connection, World is per Player,
	  Entity instances are associated with a particular World.
	  It may be possible at this point to have more than one
	  Player for a given Connection, someone should check this.

Wednesday, July 3, James Turner

	* Compile fixes for OS-X / GCC 3.1
	* Renamed Time.h/cpp to Timestamp, since case insensitive file system
	  means we look like 'time.h', which is really bad.
	* Fix include of <multimap.h> to only occur under GCC 2.9.x; I think this
	  is correct, let me know if not.

Tuesday, July 2, Ron Steinke

	* Fixes so the blocking poll handles timeouts
	  created during callbacks properly

Sunday, June 30, Ron Steinke

	* PollDefault fixes
	* Timeout::poll() and Timeout::pollAll() now return info
	  on when to poll again

Wednesday, June 26, Al Riddoch

	* Switched to new location for skstream headers.

Monday, June 21, Ron Steinke

	* Added an optional timeout argument to PollDefault::poll()

Tuesday, June 18, James Turner

	* Make dtor of Meta virtual, just to be safe

Monday, April 29, Ron Steinke

	* Added check for fd == INVALID_SOCKET in PollDefault

Sunday, June 2, James Turner

	* Safety tweaks in Connection::recv() and Metaserver::recv()
	  to handle bad sockets without segfaulting. Should
	  improve reliability of the Dime meta-server code.

Wednesday, May 22, James Turner

	* Modified Meta::got_data to be safe even when _stream is
	  NULL. This is legal when the meta-server has finished
	  sending us the server list, but there are still
	  active queries (which is quite commmon, when you think
	  about it)

Tuesday, May 21, James Turner

	* Protected against MOVE ops with no ID argument set: fail
	  rather than asserting.
	* Protected against character refreshed prior to account
	  login. (Thanks man-di)

Wednesday, May 15, James Turner

	* Added implemention of missing getServerCount to the Metaserver
	* Fixed the Meta update code to emit CompletedServerList after
	  the last active query is deleted. [Thanks to Xmp for catching
	  both issues, which no one has touched before]
	* Tweaked emission of CompletedServerList to do the right thing
	* Hopefully fixed Meta::doFailure to remove _stream from the Poll
	  object (should stop the stream of errors in Dime)
	* Converted Meta::Failure to pass it's error by const reference, 
 	  like all the other signals do

Tuesday, May 14, Al Riddoch

	* src/Player.cpp: Ensure that USERNAME is provided for logins.

Sunday, May 5, James Turner

	* Fix Connection::Disconneting system (found by unit testing!)
	* Made more of the tests work (had to stop using tcp_server_socketm
	  becuase it seems you must set SO_REUSEADDR before calling bind())
	* Fixed some return-by-reference bugs in testUtils, which again
	  helps more of the tests work.
	* Some tests are still not working right, GDB is being unhelpful
	  though. Investigation by people who's GDB doesn't suck
	  appreciated (eg, for whom 'break __throw' does the right thing)

Thursday, May 2, James Turner

    * Changed Time::Stamp::getCurrent() into a static
      factory method now(). The previous syntax looked
      really odd.
    * Updated all the callers of above.
    * Modified stubServer to set SO_REUSEADDR on it's
      sockets, which should fix the odd test problems (well some
      of them .. it still gets linker errors which is a pain).
      Thanks to elefth for spotting this one.
    * Added some boot-strapping code to TypeInfo, not sure
      how well it will work, but it uses the same path as
      loading from atlas.xml, so it should be safe.

Monday, April 29, James Turner

	* Add createRoom method to Room, to enable explicit creation
	  of OOG rooms.
	* Added a Changed signal to Room, with identical semantics to
	  Entity; this can be used to receive notifications when room
	  attributes such as name, topic and the set of child rooms
	  changes.
	* Extended Lobby to deal with room creation internally,
	  including a handler for SIGHT(CREATE) ops.
	* Made Room::getID() more robust in the case where the ID
	  is not available; now throws a meaningful exception, 
	  instead of returning an empty ID.
	* Added initial lobby tests; as usual only partial coverage,
	  but better than nothing.
	* Changed configure to simply detect CppUnit automatically,
	  so --enable-cppunit is now uncessary.
	* Updated the README and INSTALL documents a bit
	* Removed Meta::poll() and made recv() protected; these are
	  both obsolete since the new Polling system was added.

Monday, April 29, Ron Steinke

	* Wrapped calls to PollDefault::removeStream() in Meta in a check
	  that _stream->is_open()

Sunday, April 28, Ron Steinke

	* Fixed signed vs. unsigned problems in Time::Stamp by making
	  everything signed.
	* Added a getCurrent() member function equivalent to getCurrentTime(),
	  made getCurrentTime() an inline wrapper around it

Sunday, April 28, Al Riddoch

	* src/Player.cpp: Fix so that GotAllCharacters is emitted when
	  there are no characters to get.

Saturday, April 27, James Turner

	* Updated configure to use AM_PATH_WFMATH, require
	  version 0.2.6, and added the macro to acinclude
	  so people don't get broken.

Wednesday, April 23, James Turner

	* Comitting some docs I've been working on since
	  Christmas, in TeX format initially. Much work ahead,
	  contributions gladly recieved.
	* Removed the old math types code (WFmath is used by
	  all the clients now)
	* Bumped the version up to 0.9.5 now the old
	  Point / Orientation / BBox stuff is gone.
	  
Wednesday, April 17, James Turner

	* Added entity tests (incomplete, and not linking for some
	  odd reason). Something to do with WFMath I suspect, it's
	  all rsteinke's fault :)

Wednesday, April 17 2002, Ron Steinke

	* Added a test to PollDefault: if there are no file
	  descriptors, fall out immediately before trying
	  to poll

Wednesday, April 17 2002, Ron Steinke

	* Changed Time::Stamp from a typedef'd struct to a full
	  class, operator-(const Stamp&, const Stamp&) is now
	  recognized properly

Monday, April 15 2002, James Turner

	* Unit tests! Configure with --enable-cppunit, and do
	  'make check' to run them. Note they only cover a few
	  classes at present, feel free to add more, or suggest
	  them in RT.

Wednesday, April 10 2002, James Turner

	* Fixed OOG emotes in Room.cpp (dispatcher logic was totally
	  wrong, the replacement is big but correct)
	* Moved logging resposibilites from Connection to seperate
	  Log.h file. Presently, this means clients which used to
	  attach to Eris::Connection::Log should now use
	  Eris::Logged.
	* Updated all the Eris log call sites to use the new impl
	* Numerous Metaserver tweaks to make it more stable when
	  connection errors and timeouts occurs; needs more work,
	  alas.

Monday, April 08 2002, Al Riddoch

	* src/Entity.h, src/Entity.cpp: Add a flag which indicates whether
	  an entity has a bounding box.

Thursday, April 04 2002, James Turner

	* Comitted Ron Steinke's polling changes. These add a new
	  class (Poll), which is an abstract polling interface
	  by the connections. A default implementation is supplied,
	  which simple calls 'select()': to retain the current Eris
	  behaviour, simple replace Connection::poll(), Meta::poll()
	  and Timeout::pollAll() with a single call to
	  Eris::PollDefault::poll() (defined in Eris/PollDefault.h).
	  If you're using Gtk+ or some other widget set that provides
	  idle and input notifications, then you can write your
	  own Poll subclass (or steal one: Ron already has a Gtk+
	  implementation!) and everything will work.

Friday, March 15 2002, James Turner

	* Fixed logic in setContainerById, when the parent was
	  already defined: was doing 'setContainer(this)' which
	  is clearly nonsense.
	* Added in / corrected lots of debugging stuff

Tuesday, March 12 2002, James Turner

	* Fix for #222 - made logic in BaseConnection more robust
	  in handling unusual cases / states. If there are other
	  ways to 'wedge' the connections, please let me know.
	* Fix for #216 - logic in World::recvSightDelete to fix
	  container ship of visible contents
	* Preliminary work / re-factoring for #217 : added 
	  setContents helper to Entity, and handle that case
	  in setProperty.

Sunday, March 10 2002, James Turner

	* Set a default orientation on Entities, to avoid some
	  nasty behaviour that Simon was seeing. Set w=1, x/y/z=0.

Friday, March 08 2002, James Turner

	* Fixed .so version information. Thanks to Damien for the
	  motivation, and Ron Steinke for showing me how.

Thursday, March 07 2002, James Turner

	* Changed delay on select() to be 0,0 : previously it was
	  10 milliseconds, which is a bit harsh on the clients.
	* Changed World::recvSightMove to use the ID argument of
	  the Move operation, instead of it's FROM. This should
	  fix reporting of pick/drop ops. Both these fixes
	  are suggested by Al Riddoch.

Wednesday, March 06 2002, James Turner

	* Fixed character looks to actually work (had name and type
	  name args to ClaassDispatcher child in the wrong order!)
	* Implemented disptch-safe adding and removal of subdispatchers.

Monday, March 04 2002, Al Riddoch

	* Fix removal of deleted entities from their container.

Saturday, March 02 2002, James Turner

	* Applied Ron Steinke's patch for WFMath support
	* Bumped version to 0.9.2 so people can test against the
	  WFMath-enabled version
	* Added README-wfmath-conversion, explaining how to
	  migrate a client to the new API

Monday, February 11 2002, James Turner

	* Implemented Player::getAccountID();

Sunday, February 10 2002, James Turner

	* Added wait for INFO(LOGOUT) response from server, 
	  and a timeout in case the server doesn't support that
	* Added a signal on Player when a logout completes
	* Added a comment in World.cpp about the TO field on
	  IG LOOK ops.
	* Added another 'foundation' dispatcher to Connection
	  (op:info:op)

Friday, Febuary 08 2002, Al Riddoch

	* Clear TO on IG Look ops.

Sunday, February 03 2002, James Turner

	* Fixed logout (set FROM)
	* Changed character API, let me know what you think.
	* Implemented TypeInfo::getParentsAsSet

Tuesday, January 29 2002, James Turner

	* Support for sending private chat messages (using a new
	  'msg' method on Person)
	* Root-entity handling fixes on World and Entity
	* Person now caches the Lobby it's bound too (should help
	  when we eventually have multiple connections)

Thursday, January 24 2002, James Turner

	* Supports skstream2, thanks to Michael Koch for the 
	  work. 

Sunday, January 20 2002, Al Riddoch

	* Add children accessor to TypeInfo node.

Friday, January 18 2002, James Turner

	* Move lots of extraneous public stuff in TypeInfo
	  private as it should have been all along. If this
	  breaks your client code, talk to me, since you
	  were maybe doing something unwise.
	* Added global BoundType signal on TypeInfo, to
	  inform clients about new types as they become
	  available. Suggested by Al Riddoch for Equator work.

Thursday, January 17 2002, James Turner

	* Further property tweaks suggested by Karsten.
	* Added observeProperty accessor to Entity, which allows
	  easy public monitoring of attribute values.
	* Fixed connectOp[To/From]Slot on Entity
	* Added getType() helper on Entity that returns the
	  corresponding TypeInfo*

Tuesday, January 15 2002, James Turner

	* Rewrote property handling completely : should be
	  much simpler (no unsync() / resync()) and much more
	  flexible now. Properties now have get and set signals
	  that can be observed for value changes.
	* Moved all the attribute handling into setProperty(),
	  removing the coresponding code from recvSet/Sight/Move.
	  This unifies the decoding of every attribute, but
	  also bypasses the Atlas::Objects handling of core
	  attributes (only relevant for name, location, stamp
	  and position).

Monday, January 14 2002, James Turner

	* Skstream exception handling fixes, suggested by
	  Stepher Meier.
	* Fixed to bugs and general tweaking of attribute synchronisation
	  in Entity : this code hasn't been excercised before, so it's
	  a bit wrinkly.
	* CodeWarrior and GCC 3.0 return value warning fixes from
	  Michael Koch and Jesse Jones

Wednesday, December 26 2001, James Turner

	* Tweaked configure.in slightly, to use CXXFLAGS. Should
	  correctly pick up CXXFLAGS / CFLAGS variables if they are set.
	* Made an assertion in Room.cpp more verbose becuase I hit it once,
	  using stage. This maybe a sequencing issue (similar to what
	  happens when logging in before the type hierarchy has fully
	  transferred)
	* Provide a platform specific way to access the underlying connection
	  primitive (i.e the socket's file-descriptor under POSIX). This is
	  to allow efficent integration with the Gtk+ main-loop system,
	  which cna automatically poll file-descriptors and invoke callbacks.

Saturday, December 22 2001, James Turner

	* Committed build improvement by Michael Kocg - Eris now as
	  an eris-config script and M4 macro file, so much more
	  reliable configure scripts can be setup by clients.

Friday, December 21 2001, James Turner

	* Standardised Entity signals to *not* include 'this' as
	  an argument. If this behaviour is desired, using SigC::bind()
	  to achieve the same affect when connection to the Signal.
	  An example of doing this can be found in silence/src/game.cpp
	* Changed a few signals to use pass-by-reference, since this
	  is valid.
	* Calling this version 0.9 in preperation for a beta release in
	  a few weeks.

Friday, December 14 2001, James Turner

	* Correct processing of 'face' attr : it's a unit vector giving
	  the direction, not a heading float. Thanks to Karsten for the
	  info, hopefully this should work now.
	  
Wednesday, December 12 2001, James Turner

	* Added Quaternion class to Types.h and Types.cpp
	* Added _orientation member to Entity, an accessor and code in
	  recvMode to process it.
	* Added a temporary check for the Acorn 0.3/0.4 'face' attribute,
	  and map it to a quaternion if the orientation attr is not found.
	  This can be removed in a few weeks, but takes the pressure off
	  alriddoch to update cyphesis-C++ 'right now'.
	  
	  All of the above is untested becuase I don't have a graphical client
	  to test with; it compiles fine, and it's pretty simple, but
	  beyond that I'm not sure.

	* Moved an assert in Room.cpp (line 244) whcih was a bit premature
	
Wednesday, December 12 2001, James Turner

	* Fixed OOG recvAppearance/Disappearance to handle multiple
	  entities in ARGS. (RT ticket #165). I have verified that the current
	  behaviour still works, can't test the actual multi-args becuase
	  no server sends that format (I think)
	* Added getVelocity accessor to Entity, which somehow got left out
	* Updated Entity::recvMove to update the velocity value.

Thursday, December 06 2001, Al Riddoch
	* Strict C++ compliance fixes, tested with gcc3.

Thursday, November 29 2001, Al Riddoch
	* src/atlas_utils.h: Avoid copying twice when casting.
	* src/Utils.cpp, src/Utils.h: Return const objects.
	* src/Types.cpp, src/Types.h, src/Entity.cpp: Remove support for
	  bmedian attribute now deprecated.

Thursday, November 29 2001, James Turner

	* Changed to actually call World::mark[Inv|V]isible from
	  Entity::setVisible; thus the IEC gets some testing.
	* Handle various edge-cases in the IEC correctly, notably
	  when there are no buckets defined
	* Various comment and dead-code clean-ups.

Monday, November 26 2001, James Turner
	* Re-built dispatch system with various new features:
	* Dispatcher is now abstract, since much logic has been factored
	  out in the StdBranchDispatcher
	* Split Encap dispatcher apart so it only performs de-encapsulation,
	  no type/class selection
	* Class dispatcher now selects the most specialised binding for
	  a type; i.e if handlers for 'SIGHT' and 'INFO' exist, only the
	  one for 'SIGHT' will be fired.
	* Updated all the users of dispatchers to use the new structure
	  and interfaces - tested and most things seem to work, including
	  IG entry. OOG chat emotes are broken, will fix tomorrow.

	* Handle Bounding boxes, the 'BMEDIAN' attribute and velocity/
	  position better in Entity.
	* Added BBox to Types
	
Saturday, November 24 2001, James Turner

	* Cleaned up logging output for TypeInfo.cpp
	* Extended World to provide a protected mark[Inv/V]isible interface to
	  entities, so they can control their placement in the invisible entity
	  cache.
	* Added InvisibleEntityCache.[cpp/h] and hooked it into World.cpp
	* Added World::tick() method to do periodic updates (eventually motion
	  prediction and so on), initally just flusing the InvisibleEntityCache
	  as necessary.
	* Updated Time to include an operator- as well as operator+
	* Logs in flawlessly with an up-to-date build of Cyphesis-C++!

Thursday, November 22 2001, James Turner

	* Lots of TypeInfo changes, basically the whole dependancy system
	  has been gone over to make it much more robust. There is one
	  (known) issue remaining, but a large number of other issues have
	  been fixed, notably that logging in with a current version of
	  Cyphesis-C++ should work
	* Added a 'listUnbound' debugging helper to TypeInfo
	
	* Add a check in Lobby's ctor if the connection is already active;
	  if so, run the netConnected callback directly. This avoids
	  incorrect behaviour in certain intialisations orders.
	* Rewrote World::recvAppear to deal with every argument in the
	  list, not just the first one. 
	* Added isVisible and setVisible methods to Entity, which are
	  driven off Appearance and Disappearance ops from the server

Sunday, November 18 2001, James Turner

	* Changed objectSummary to special case handling of the Root/meta
	  object (don't report it as <invalid>, it isn't!)
	* Made atlas_cast return const, as a matter of policy against
	  accidental evil
	* Login failures now clear the current action / serial in Player
	  after reporting; thus repeated login attempts now actually work!
	  (it's really true this time, I promise : I've actually tried it
	  and it works)
	* Move the serialno validator around so it A) only operates on
	  messages recieved from Atlas (don't worry about redisaptches /
	  internal crap) and B) correctly handles encapsulation on the
	  dispatch stack (LeafDispatcher sets and ObjectArrived checks
	  the back (top) of the stack, not the front)
	* Re-wrote 'Connection::clearSignalledWaits' to use a manual iteration,
	  since I am clearly doing something stupid with STL remove_if; it was
	  leaving some objects in the queue. My reading for tonight is
	  Scott Meyer's 'Effective STL', <sigh>
	  
	* Added a 'LogLevel' argument to Eris::Log, and changed all the callers to
	  set it a sensible value; should reduce the amout of crap we have to
	  wade through.
	* Changed atlas_cast to use AsObject().AsMap() since AsMap doesn't
	  work correctly for RootEntity/Operation : this was causing lots of
	  strange pseudo-corruption bugs. Eternal gratitude and your first
	  born to be sacrificed to the greater glory of alriddoch!

Thursday, November 15 2001, James Turner

	* Connection::reconnect() now generates a 'Failure' signals instead of
	  throwing InvalidOperation; I was being a tad aggressive here for no
	  good reason.

Saturday, November 10 2001, James Turner

	* Made TypeInfo::init safe (doesn't assert) during reconnections
	* Updated eris.dox.in with better settings
	* Added more documentation to classes and generally improved the
	  quality and accuracy, still a long way to go but getting better
	* Made objectSummary much more robust handling suspicious objects,
	  including a try { .. } catch block for the Atlas exceptions,
	  and more aggresive IsXXX checking on the Message::Objects.
	* Similarly made the getMember/Arg helpers check the structure of
	  the input objects before extracting the data. This is slower but
	  all done inside asserts()
	* Changed the access control for various methods (eg Room's ctor) as
	  a result of looking at the generated documentation, to avoid
	  exposing lots of internal stuff to the client.

Thursday, November 08 2001, James Turner

	* Added LoginFailure signal to Player, which is generated instead
	  of an exception being thrown in response to server-side failures
	  in account creation / login.
	* Updated Player::login/createAccount to work correctly with
	  multiple attempts / tries. This involved moving the 'Connection'
	  argument to both up to the Player constructor, where it should
	  probably have been all along.

Tuesday, November 06 2001, James Turner

	* TypeInfo stuff basically works, parses the atlas.xml file
	  if it's found in the same directory as the program.
	* Changed the way INFOs are handled since we have proper
	  inheritance now
	* Various other fixes which needed type data to be working, 
	  e.g. the dispatchers for OOG now bind (correctly) to 'account',
	  not 'player', so admin accounts and so on will work.
	* Probably lots of bugs lying in the grass, but it works okay
	  and I want to get this commited.

Sunday, October 28 2001, James Turner

	* Made dispatchers ref-counted, just like Python and COM. This is
	  becuase the dispatcher tree turned into a dispatcher graph, so
	  ownership got more complex.
	* Preliminary work on ClassDispatcher to support TypeInfo;
	  added ClassDispatcher.cpp (not compiled at present) which uses
	  the TypeInfo functions. The current behaviour is retained, since
	  it's faster and common (at least for operations which make up the
	  bulk of the decode traffic), and because it's necssary to
	  boot-strap the type data transfer.
	* Lots of changes to the TypeInfo code so it might actually work:
	* Made all those dumb functions into static methods
	* Added preliminary support for integer typeids which are coming
	  with Atlas 0.5.x and will make the world a happier place
	* Moved the 'Bound' signal into TypeInfo which is cleaner and gets
	  rid of a global table
	* Still much to do before it compiles or works
	  
Thursday, October 25 2001, James Turner

	* Tweaked objectSummary to handle SET ops better, still needs more
	  work
	* Changed the SET op synthesised by Entity::setContainerById to
	  include the encapsulatig SIGHT : this means the SET actually gets
	  processed insetad of just being ignored
	* In Connection debug mode, store pairs of FROM/SERIALNO for
	  recieved ops, and ensure their uniqueness. Ops with SERIALNO==0 are
	  flagged with a warning, this needs to be clarified. This is
	  correctly flagging the duplicate recieve bug, w00t!
	* Re-wrote Connection::clearSignalledWaits to actually work
	* Made ~WaitForBase virtual so delete doesn't segfault, d'oh
	* Added WaitForBase as a friend of Connection so it can call the new method
	  addWait which actually puts the WaitFors into the WaitList; this wasn't
	  happending, hence the bug with WaitFors being fired multiple times.
	  
	* Added LeafDispatcher, a base class for (guess what?) leaf dispatchers,
	  that sets a special attribute (__DISPATCHED__) when it's dispatch()
	  method is invoked
	* Connection::ObjectArrived tests (in debug mode) that all messages 
	  have their __DISPATCHED__ attr set after being dispatched (unless an
	  exception is thrown). This provides a warning in the log files if
	  a message is completely ignored and 'drops out the bottom' of the
	  disaptcher tree
	* Updated SignalDisaptcher0/1/2 to inherit from LeafDispatcher and
	  call it's dispatch implementation.
	  
	* Changed World::recvSightCreate to just call recvSightObject with
	  a synthetic SIGHT op; the possiblity to do special CREATE only
	  processing still exists, but this avoids problems where
	  SIGHT(CREATE(entity)), LOOK(entity) and SIGHT(entity) get overlapped
	  and so on.
	  
Tuesday, October 23 2001, James Turner

	* Added objectSummary function in Utils, which gives a compact
	  string summary of Atlas operations and entities; the heuristics
	  need some work but the output is usable (in Eris::Log calls for
	  examples)
	* Added a '_debug' option to Connection (always set to true at
	  present) that controls some Logging output and Atlas send/recv
	  logging
	* Renamed the Atlas log files to make their function more obvious
	* Bullet-proofing in World::RecvSight to deal with multiple sights
	  and unrequested sights of entities.

Wednesday, October 17 2001, James Turner

	* Fixed Lobby::join to handle re-joins and actually return
	  the Room instance.
	* Added PrivateTalk signal to Lobby and the necesary code
	  to emit it.
	* Adjusted the inital entry logic for rooms to ensure the
	  Entered signal is always emitted correctly.

Tuesday, October 16 2001, James Turner <jmtn@blueyonder.co.uk>

	* Made Room::say/emote set the loc arg correctly; Eris now works with
	  both cyphesis and stage, so no one can complain.

Monday, October 15 2001

	* Fixed up dispatcher tree some more (sights encap imags), seems
	  to work with Stage, and Stage works with Process. This is
	  tenuous validation at best, but it will do. Testing with
	  cyphesis appreciated.

Sunday, October 14 2001, James Turner

	* Added ArgumentDispatcher (map args only,  list args are easy too add too).
	  This is required to route based upon args[loc=ID] in the new OOG
	  format we are switching too.
	* Rebuilt the Room and Lobby OOG routing code for sight / sound / imaginary
	  apperance / disappearance. The code has been flipped; Lobby sets up 
	  global class selectors for each op, and rooms insert ArgDispatch
	  children below these to examine the 'loc' attribute.
	* This needs some testing and verification, but I can't right now so
	  comitting and hoping.

Monday, October 01 2001, James Turner

	* Support deffered binding of containers in Entity::recvSight,
	  using the same policy as 'set' uses; i.e creating a secondary
	  Set operation and deffering it until after the container is
	  loaded.
	* Check for empty location attributes in Entity::recvSight, which
	  can happen with root / world objects.
	* Changed various instances of GetAttr("foo").AsXXX to GetFoo() where
	  I'd forgotten / didn't know a Foo accessor existed; thanks to
	  alriddoch for catching these.
	* Fix emission of World.Entered signal until both the root entity and
	  character entity are valid.
	* throw an InvalidOperation exception if getRootEntity is called
	  prior to World.Entered being emitted.

Thu, 27  Sep  2001 13:50:42 +0100, James Turner

	* Changed inheritance of BaseException; now inherits from std::runtime_error,
	  and correctly forwards the error string. As a result, exception.what() will
	  return a sensible value. Thanks to Karsten Laux for the suggestion.
	  
	* Applied Karsten's re-ordering in World::World() (store the instance pointer
	  before creating the default entity). I have no idea why this  went
	  un-noticed in testing, very strange.
	* Added an id member to BaseConnection, to help discriminate connections and
	  their timeouts. This is important if you're connecting to a server while
	  the Meta code is querying it : things get confused!. Thanks to Karsten for
	  the catch on this one.

Sunday, September 16 2001, James Turner <jmtn@blueyonder.co.uk>
	* Correclty process the location / contains
	  attributes to construct the IG entity tree

Monday, September 10 2001, James Turner <jmtn@blueyonder.co.uk>
	* Handle both 'account' and 'player' entites in Lobby;
	  necessary until proper client-side Atlas type handling
	  is available.
	* Configuration changes to lay ground work for Python
	  bingings - talk to me if you want to help / test

Wednesday, August 29 2001, James Turner <jmtn@blueyonder.co.uk>

	* Fixed disconnection (actually call _stream->close())
	  and hence reconnection - rejoice!
	* Added a simple logging system using va_args and a
	  signal on Connection (Log). Seems to work, and doesn't mess up
	  ncurses like simply using 'cout' does. Wish I'd done this ages
	  ago, took about 5 minutes <sigh>.

Tuesday, August 28 2001, James Turner <jmtn@blueyonder.co.uk>

	* Added #include "ServerInfo.h" to Metaserver.h so people don't
	  get hit with a huge raft of STL errors if they forget to include
	  it themselves.
	* Added timeouts to Metaserver and hooked up MetaQuery timeouts
	* Fixed various disconnect / reconnect bugs
	* Player now hooks Connection::Disconnecting to issue a logout
	  operation. Currently the reply is undefined so this will always
	  hit the disconnect timeout.
	* Disconnect is still not working quite right.

Monday, August 27 2001, Jmaes Turner <jmtn@blueyonder.co.uk>

	* Add timeouts to BaseConnection / Connection, and
	  exposed in a relatively sane way.
	* Modifed Time.h/.cpp to be more portable, and
	  updated the timing code in Meta that used it.
	* Made Connection::disconnect actually do something. This entailed
	  adding status locking to the connection =>
	* Added lock() / unlock() calls to Connection; these can hold the
	  Connection in a state (currently only DISCONNECTING) while
	  users of the connection perform tasks. Needs testing of
	  course, but I think the principle is sound.
	
Sunday, August 19 2001, James Turner <jmtn@blueyonder.co.uk>

	* Added support for emotes to OOG session
	* New TypeInfo system (not tested yet)
	* Re-wrote the Connection dispatch model to handle 
	  re-dispathces more flexibly.
	* Added a zero-argument SignalDispatcher
	* Fixed Room Talk and Emote signals to send the
	  account ID, instead of the user's name.

Friday, August 10 2001, James Turner <jmtn@blueyonder.co.uk>

	* Updated README and INSTALL to contains something
	  vaugely informative.
	* Initial work on TypeInfo.cpp/.h classes (not part)
	  of build at present.
	* Entity factories are sorted by a priority now. This
	  meant using a multimap internally, though I've just
	  realised a multi-set would be sufficent. D'oh.

Saturday, August 04 2001, James Turner

	* Handle sparse rooms without throwing an exceptions;
	  checks that the 'people' and 'rooms' list exist
	  before attempting to extract them.
	* Fixed file line endings

Sunday, July 29 2001, James Turner

	* Added redispatch (re-post) support; needs testing
	  and documentation but the basics are there and
	  quite clean.
	* Added RecapDsiaptcher and IdDispatcher
	* Instance() singleton accessor on Eris::Connection
	* Sythetic appearance generation as required
	* Fixed dispatcher setup for sound(talk)
	* In-game works, with some bug (5 copies of messages!)

Friday, July 27 2001, James Turner

	* Renamed methods to leadingLowerCase style, as requested
	  by Al Riddoch.
	* Added std:: prefix to lots of things to make GCC 3.0
	  happier. Doubtless some remain.
	* Fixed context dispatch stack handling when backtracking
	  occurs (wasn't correclty popping encapsulated objects)

Thursday, July 26 2001, James Turner

	* Added Meta-server status codes
	* Added 'Failure' signal to Meta
	* Much improved meta-server error handling in general

Wednesday, July 25 2001, James Turner

	* Changed Dispatcher interface to support passing
 	  multiple objects down the tree. This enables 'context'
	  to be maintained, i.e encapsualting objects are not
	  discarded.
	* Added SignalDispatcher2, which binds to both the curent
	  top of the context stack, and it's parent. In theorey,
	  SignalDispatcher3/4 etc could be provided, but this is
	  unlikely to occurr much in practice.
	* Bumped version to 0.1.3
	* Handle Stage RIM Chat bug where ID of certain objects 
	  is not set in SIGHTs (use TO attribute of SIGHT op)

Sunday, July 22 2001, James Turner

	* Fixed player to correctly register 'LoggedIn' signal
	  on Lobby, which was getting ignored.
	* Implemented Connection::RemoveDispatcherByPath

Saturday, July 21 2001, James Turner

	* Fixed Lobby::Join to match silence-py behaviour
	* Added 'Leave' (part) operation to Room, and made the
	  destructor call it when appropriate.
	* Hacking charatcer creation, still being strange

Sunday, July 15 2001, James Turner

	* Meta-server code now seems to work, but:
	* Rebuilt the query code to build and return ServerInfo objects
	* Handle Cyphesis-style anonymous GET replies (no refno set)

Wednesday, July 04 2001, James Turner

	* More reconnection work
	* Correctly handle NULL stamps (assume unstamped)
	* Bumped version to 0.1.2

Tuesday, July 03 2001, James Turner

	* Added reconnection support to Connection
	* Changes to World to support reconnection
	* Correctly validate the entity stamp when processing
		APPEARANCE ops.
	* Added GetStamp() accessor to Entity
	* Set serial number on most out-going ops	

Monday, July 02 2001, James Turner

	* Added meta-system; needs some work

Wednesday, June 27 2001, James Turner
	
	* Re-factored connection into an underlying layer (BasicConnection)
	and a public wrapper (Connection); this is to permit Meta-system
	queries using a light-weight connection system.
	* Add Appear / Disappear / Entered signals to World

Tuesday, June 26 2001, James Turner

	* Updated configure.in to test for skstream header / library
	* Updated configure.in to test for Atlas headers and libraries
	* OOG System appears to be working with preliminary code in Stage
	* Appear / Disappear handled correctly
	* Generates correct OOG Talk ops
	* Preliminary support for IG operation
	
Friday, June 15 2001, James Turner
Created initial change-log becuase automake requires it