EPIC4pre2

*** News 06/18/99 -- New flag to $info, $info(v)
	The $info(v) function call returns the "class major minor" version
	information.  For example, when this item was written, $info(v) was
	to return "epic4 pre2 004", indicating an epic4pre2.004 class client.
	Other clients may very well use $info(v) to return version information,
	and hopefully they wont use 'epic4' as the 'class'. =)

	The reason this is being added is because technically speaking, 
	looking at $J for the version info is a bad idea, and $V is good,
	but $V overlaps for all the different classes of ircII clients.
	So if (word(0 $info(v)) == [epic4]) then you know youre using an
	epic4 client.
	
*** News 05/28/99 -- New built in function, $ttyname()
	It just calls the ttyname(3) function, which returns the name of
	the tty that the epic process is attached to, if for whatever
	reason you care about that.

*** News 05/28/99 -- New built in function, $lastlog()
	This function roughly simulates the /lastlog command, except it
	returns a list of numbers that you can pass to $line() to get 
	interesting information from the lastlog.   The syntax is:
		$lastlog("window desc" "pattern" "lastlog level")
	There must be exactly three arguments passed to this function, and
	so if any of them contain spaces, you will need to use double
	quotes to protect the wording.  The function will return all lines
	in the lastlog buffer of the specified "window" that match "pattern" 
	and also are of one of the types contained in "lastlog level".  

	For example:
		$lastlog(0 *foo* ALL)
	Returns all of the lines that contain the string 'foo' of any
	lastlog level type in the current window.

	Important note #1: Lastlog line numbers are counted from the bottom 
	of the display to the top.  The most recently displayed line is always
	numbered 0.  The line numbers are also not static.  Lines are 
	continually being added and deleted for each new line of output.
	This function returns the line numbers in numeric order, which means
	from the bottom up.  If you want to traverse from the top down, you
	will need to pass the return value of this function through $revw()
	to get them in the desired order.

	Important note #2:  Because 'echo'ing to the screen changes the 
	order of the line numbers, extreme caution must be used when 
	outputting anything based on the $lastlog() command -- if you are
	not careful to compensate, you might just retreive the wrong line.
	It is _not_ my fault if this happens.

*** News 05/28/99 -- Two new extensions to the $line() function:
	First, you may now pass a window name or refnum as the optional
	second argument.  Previously, you could only pass in a window refnum.
	Second, you may also specify the literal word "-LEVEL", and the
	return value of $line() would include the lastlog level of the
	requested line as the first word:  For example:
		$line(0 -LEVEL)
	Might return
		"CRAP This is some text"

*** News 05/07/99 -- New function, $realpath()
	If your system provides the realpath(3) library call, then this 
	function will convert a relative or absolute pathname into a
	canonical pathname.  All symlinks, and any '..' or '.'s are 
	handled.  Note that *THE DIRECTORY PART OF THE PATH YOU PASS TO THIS
	FUNCTION MUST ACTUALLY EXIST*, because this is how the library
	call works.  If an error occurs (such as the path not existing), 
	the false value is returned.  If the realpath library call is not 
	available on a system, then this function will simply return the 
	arguments as-is.

*** News 05/07/99 -- New scripts, 'dig' and 'debug.irc'
	Refer to them for more information.

*** News 05/07/99 -- New function, $indextoword()
	This function converts an $index()-type return value to a $word()-type
	argument, so that you can extract the entire word from a string that
	contains a character that you know about.  For example:
		$word($indextoword($index(! $*)))
	would return the entire word that contained the first instance of 
	the bang character in $*.  This might be useful to extract words 
	from the input line, when used with $curpos() and $L.

*** News 05/07/99 -- 20 new /set status_user<n>'s.
	There are now 20 new /set status_user's available, for a grand total
	of 40.  The new values are /set status_user20 through status_user39.
	They show up on your status bar as %{2}0 through %{3}9.  The 
	difference between these 20 and the other 20 is that these always 
	display on all your windows, not just the current-type window.

*** News 04/08/99 -- New feature to aliasctl, $aliasctl(* exists *)
	This new verb to aliasctl returns 1 if the indicated alias/assign
	exists and 0 if it does not.

*** News 02/01/99 -- New operator in new math paser, <=>
	If you are a user of the new math parser (see below, /xdebug new_math)
	then there is a new operator available.  The <=> operator is a binary
	operator that takes two lvalues and swaps the rvalues of each operand.
	For example, if $foobar is 1 and $booya is 2, then
			@ foobar <=> booya
	Results in $foobar being 2 and $booya being 1.  The operator is right
	to left associative, which permits you to string them together to
	get a "shift rvalues to the right" operation.

*** News 02/01/99 -- New built in /set, /SET NEW_SERVER_LASTLOG_LEVEL
	Some users do not appreciate that when you do a /window server
	operation, that the lastlog level of the window is changed to
	LOG_ALL.  So this /set now controls what it will be set to.  The
	default is (obviously) LOG_ALL.  Some of you will probably change
	this to LOG_NONE, I suppose.

*** News 02/01/99 -- New built in flood/ignore levels	(srfrog)
	There are three new flood-protection types, and three new ignore
	levels:
		"TOPIC"		"JOIN"		"NICK"
	respectively covering topic changes, channel joins, and nickname
	changes.  

*** News 02/01/99 -- New built in /set, /SET RANDOM_LOCAL_PORTS
	For the delusionally paranoid, you can /set this to ON if you
	want EPIC to make a token attempt to use a random port for each
	outbound DCC request.  If EPIC randomly picks a port that is in
	use, then as a fallback, it will just ask the operating system
	for a port (to avoid resource starvation).

*** News 02/01/99 -- New built in /set, /SET RANDOM_SOURCE
	EPIC now has more flexibility in what it uses as its random
	number source.  You may now choose between three random number
	sources by setting this /set to one of the following values:

		0 - Use the /dev/urandom device
		    PRO: Reasonably good source of non-repeating numbers.
		    PRO: Reasonably good source of entropic numbers.
		    CON: Not widely implemented outside of linux/freebsd

		1 - Use the classic ircII psuedo-random number generator
		    PRO: Very good source of non-repeating numbers.
		    PRO: Always available
		    CON: Numbers generated have almost no entropy.

		2 - Use the alternative epic gettimeofday() generator
		    PRO: Almost always available
		    PRO: Numbers generated have modest entropy
		    CON: Poor source of non-repeating numbers.

	If you /set this to 0 or 2, and the appropriate mechanism is
	not available, then it will silently use the built in psuedo-random
	number generator, which is always available.  This variable is used
	by the global random number generator, which is used by the DCC
	random port guesser (see above), the $random() built in function,
	and the $randread() built in function.

*** News 02/01/99 -- New built in function, $hash_32bit()	(srfrog)
	This function takes a case-insensitive string as its first argument 
	(so if it has spaces, you need to surround it with double-quotes) and 
	an optional argument specifying the number of characters to hash, and
	converts that number of characters in the string into a 32 bit integer.
	The 32 bit hash result is returned.

*** News 01/25/99 -- New feature, alias argument lists
	This is kind of hard to explain, and I implemented it all in one day,
	so if you try real hard to break this, you probably will.  You may 
	now supply argument lists to your aliases.  An alias list is one or
	more local variables that will be given portions of the arguments
	passed to your alias.  For example:
		/alias a1 (var1, var2) {echo $var1, $var2)
	then
		/a1 one two
	$var1 becomes "one" and $var2 becomes "two".  $* is the empty string.

-	If the user passes more arguments than you have on your list, then
	all of the excess arguments will go to the last variable.  So in
	the above example, if you do:
		/a1 one two buckle my shoe
	then $var1 becomes "one" and $var2 becomes "two buckle my shoe".
	$* is the empty string.

-	If the user passes less arguments than you have on your list, then
	all of the excess variables will recieve the empty value.  So in
	the above example, if you do:
		/a1 one
	then $var1 becomes "one", and $var2 is the empty string and $* is
	the empty string.

-	If you would rather have the variables in the list just "shift off"
	the first few variables and leave the rest of the arguments in $*,
	then you can specify your last variable as "...", which will just 
	pop off as many words as neccesary and leave the rest in $*.  So
	in this example:
		/alias a1 (var1, var2, ...) {echo $var1, $var2)
	then
		/a1 one two buckle my shoe
	$var1 is "one", $var2 is "two", and $* is "buckle my shoe".

-	If you would like to have any extra arguments thrown away, then you
	can specify the last variable as "void".  In this example:
		/alias a1 (var1, var2, void) {echo $var1, $var2)
	then
		/a1 one two buckle my shoe
	$var1 is "one", $var2 is "two", and $* is the empty string.

-	If you would like to provide a default value for a variable, just
	in case there is not an argument that should be there, you can 
	specify it with the "default" qualifier. 
		/alias a1 (var1, var2 default "baz") {...}
	So if you do:
		/a1 one
	Then $var1 is "one", and $var2 is "baz" (because it is the default
	value you specified) and $* is the empty string.  The default value
	is limited to one extended word, so if you want to have spaces in it,
	you need to surround it with double-quotes.  The default value will
	be expanded, so you may have expandoes (such as $C) in it.

*** News 01/13/99 -- New built in /set, /SET TERM_DOES_BOLD_BLINK (knghtbrd)
	If your terminal emulator does bold background colors when you
	set the blinking attribute, you can /set this to ON to give a hint
	to the client that it is ok to output the BLINK attribute in places
	where it would otherwise not do so.  A specific example is for the
	^C background colors 00-15, of which 4 are "bold".  If this /set is
	OFF, those 4 colors are always normal intensity, and if this /set is
	ON, then those 4 colors will be "blinked", and hopefully your emulator
	will turn that into a bold background color.

*** News 12/17/98 -- Built in command /cycle repealed
	The built in command /cycle has been repealed.  The restriction that
	required its implementation has also been repealed.  You may now do
	/quote part, and so you can use your old /cycle alias.

*** News 12/17/98 -- New built in function, $functioncall()
	This built in function can be used to determine if your user alias
	was called as a command (eg, /foobar) or as a function (eg, $foobar())
	If your user alias was called as a function, $functioncall() will
	return 1.  Otherwise, it will return 0.

*** News 12/17/98 -- Slight change to built in function argument passing
	Historically, built in functions have returned the false value 
	when a required argument was considered to be "missing".  However,
	a single space, delimited by quotation marks, has also been considered
	to be "missing", so that something like:

		$strip(" " one two three)

	returned nothing, because the required first argument was "missing".
	This has been corrected, so that quote-space-quote is now a valid
	argument that is not "missing", and that argument contains a single
	space in it.  This permits you to pass a space as a required argument
	to any function by enclosing it in double quotes.  Of course, this
	only impacts functions that interpret their arguments as a list of
	words.

*** News 12/11/98 -- New window action, /window refnum_or_swap
	It has been pointed out that there is no atomic way to make a
	given window both current and visible, without failure.  This
	action provides that capability.  It is passed one argument, the
	refnum or name of the window to be made current.  If the window is
	visible, then the client makes that window current (as /window refnum
	would do).  If the window is hidden, then the client swaps the
	window in and makes it current (as /window swap would do).  This
	action never fails unless you specify a window that doesn't exist.

*** News 12/11/98 -- New built in command, /usrip
	EFNet has decided to re-implement the /userip command from
	undernet and give it a gratuitously different name.  Ugh.
	See /userip for more information.

*** News 12/11/98 -- Status line expando %S changed, always displays
	Due to overwhelming request, the %S status-line expando now always
	displays the server information, even if you are not on multiple
	servers.  If you have a severe problem with this change, please
	let me know.

*** News 12/11/98 -- Clarification of $winchan() and $chanwin()
	Nuke has asked me to officially clarify that if you specify a
	channel argument, a second argument is optional.  The optional
	second argument is the server refnum from which you are interested
	in the channel.  This is to clarify between two channels of the
	same name on two different servers, obviously.

*** News 11/10/98 -- /on switch_windows extended a bit
	I am documenting this here because there have been some changes 
	(improvements) which contradict the previous documentation for 
	/on switch_windows, so a record is needed.  Anyhow, as far as i 
	can tell, the client now will hook /on switch_windows _every_ time
	the current window is changed, not just in some select situations.
	Hopefully this will make everyone happy. =)

*** News 11/03/98 -- $servername(), $servernick(), $serverports() changed
	They are now all happy if you do not provide them with any arguments,
	as they will default to the current server refnum (from_server).

*** News 11/03/98 -- New built in function, $prefix(<word list>)
	The $prefix() function purports to return the longest string that
	is a common leading substring to all the words in the given word
	list.  For example:

		$prefix(one two three)		returns the false value
		$prefix(p1 p2 p3)		returns "p"
		$prefix(one-1 one-2 one-3)	returns "one-"

*** News 11/03/98 -- New built in function, $maxlen(<word list>)
	The $maxlen() function purports to return the length of the longest
	string in the given word list.  You figure it out. ;-)

*** News 11/03/98 -- New verb for /window, /window clear
	This is a pretty obvious one -- how did it get overlooked?
	Pretty simple.  It clears the current window.

*** News 09/28/98 -- /on dcc_lost hooked for auto-dcc reject
	When the client auto-closes your dcc's because they have excceded
	the timeout specified in /set dcc_timeout, the client will hook
	/on dcc_lost for you with the pertinent information.

*** News 09/28/98 -- New verb for $aliasctl(), pmatch (Colten, nutbar)
	You may now do pattern matching on your aliases and assigns
	by specifying "pmatch" instead of "match".  Unlike the "match"
	on bitchx, this does not auto-append a *, so if you want "match"
	like behavior, remember to put it on the end.  Examples:

		$aliasctl(alias pmatch *foo)
			returns all alias names that *end* with "foo"
		$aliasctl(assign pmatch *g*)
			returns all assign names that contain a "g".

*** News 09/28/98 -- New built in function, $isaway() (nutbar)
	You give it a server refnum it tells you if youre away on that
	server.  The default argument is the current server.  The 
	"special" argument of -2 returns 1 if you are away on any server,
	0 if you are "here" on every server.

*** News 09/28/98 -- New flags to /exec, -line, -error, -end
	These flags allow you to specify code that will be executed for
	any of the output events from an /exec'd command.  For example:

		exec -line {
			echo stdout: $*
		} -error {
			echo stderr: $*
		} -end {
			echo ALL DONE!
		} ls

	Note that /exec -end just does a /wait %proc -cmd for you (really),
	so it doesnt save you any time to do it that way.

*** News 09/28/98 -- Semantic change to local variables
	Do to exceeding popular request, all local variables are "owned"
	by their closest enclosing atomic scope and NOT by the closest
	enclosing syntactic scope.  The /bless command does not affect
	this rule in any way.  The following illustrates the change:

		alias foobar 
		{
			local a 4
			if (1) {
				# a is 4 here
				local b 5
			}
			# b is still around here, with the value 5.
			echo [$a] [$b]
		}

*** News 09/28/98 -- New built in function, $getopt()   (Crackbaby)
 $getopt(<optopt var> <optarg var> <opt string> <argument list>)

 Processes a list of switches and args.  Returns one switch char each time
 it's called, sets $<optopt var> to the char, and sets $<optarg var> to the
 value of the next argument if the switch needs one.

 Syntax for <opt string> and <argument list> should be identical to
 getopt(3).  A ':' in <opt string> is a required argument, and a "::" is an
 optional one.  A '-' by itself in <argument list> is a null argument, and
 switch parsing stops after a "--"

 If a switch requires an argument, but the argument is missing, $getopt()
 returns a '-'  If a switch is given which isn't in the opt string, the
 function returns a '!'  $<optopt var> is still set in both cases.

 Example usage:
 while (option = getopt(optopt optarg "ab:c:" $*)) {
      switch ($option) {
              (a) {echo * option "$optopt" used}
              (b) {echo * option "$optopt" used - $optarg}
              (c) {echo * option "$optopt" used - $optarg}
              (!) {echo * option "$optopt" is an invalid option}
              (-) {echo * option "$optopt" is missing an argument}
      }
 }

*** News 09/28/98 -- Lots of changes with key bindings and /bind command
	* You no longer have to do this:
		/bind ^[ meta1
		/bind meta1-[ meta2
		/bind meta2-1 metaN
		...
	  You can now do:
		/bind ^[ meta1
		/bind ^[[ meta2
		/bind ^[[1 metaN
	  And the client wont whine at you.

	* /set meta_maps controls the number of META states you have, there
	  is no limit, but there is a floor of 5 meta maps (meta0 through 
	  meta4).  Yes, you can have as many or as few as you want! =)
	  Note that decreasing this value will destroy any bindings that
	  depend on states that are going away, the client auto-garbage
	  collects them and theyre gone forever.  Just be forewarned.

	* You no longer have to do the first thing, you can just do:
		/bind ^[[11~ scroll_back
	  The client will automagically bind all the intermediate steps to
	  open meta states and you dont have to worry about negotiating
	  the meta maps (as long as you use this all the time, the client
	  will not get confused, but if you start messing around, you might
	  lose some of your binding chains.)

	* /bind accepts a special argument, '-default' which is like /dump
	  for your keybindings, but it restores the defaults so you can 
	  continue to type. =)

*** News 09/28/98 -- Add per-window status formats
	Each window may now have its own, independant status_format variable
	seperate from the global defaults.  Theyre just like the normal status
	formats, except you access them through the /window command:
		/window status_format "..."
		/window status_format1 "..."
		/window status_format2 "..."
	It is recommended that you put double-quotes around the status format
	as it probably contains spaces and you will want it parsed as just
	one argument.  Kanan has bugged me for this feature for years, so
	hopefully he wont be the only one using it. ;-)

*** News 09/28/98 -- New flag to /xquote, -ALL (Colten)
	With /xquote -all, you can send something to every connected server,
	perhaps client-side away handling...

*** News 09/28/98 -- $aliasctl(assign match) added.  (Colten)
	You can now match variables in addition to aliases.

*** News 09/28/98 -- New expression parser (IN DEVELOPMENT)
	There is a new expression parser you can play around with by doing
	/xdebug new_math.  Note that this /xdebug will probably be retained
	in the future and will be a no-op, so its safe to use in your scripts.
	More about the new expression parser when its closer to being finished.

*** News 09/28/98 -- New set, /set max_reconnect
	Supposedly, this limits the total number of times that you reconnect
	to a server ever before it gives up and drops you down to the "Use
	/SERVER to connect to a server" error.  This has not been well
	tested, so your milage may vary.

*** News 09/28/98 -- New flag to /xecho, -x
	This flag allows you to output something with all highlight attributes
	preserved, no attributes are honored.  Note that if you repaint the
	screen, these highlights will probably magically go away. ;-)  This is
	NOT a bug, so its not as useful as it might appear.

*** News 09/28/98 -- New /on, /on switch_windows
	This is hooked whenever the client changes what it thinks your
	"current" window is in any context:
		$0	The old screen number
		$1	The old window refnum
		$2	The new screen number
		$3	The new window refnum
	Note that the window might change through forces other than the
	'SWITCH_WINDOWS' keybinding.  You can have a pretty good idea
	when this happens because $0 and $2 will be different.

*** News 09/28/98 -- New built in keybinding, /bind <X> refresh_status
	This just updates all your status lines, and nothing else.

*** News 09/28/98 -- You may now 'return' out of /fe, /do, /for, etc
	This was not supported before.  Probably a bug.

*** News 09/28/98 -- New set, /set mangle_logfiles
	You may now mangle the output that is destined to your logfiles
	by /set'ting this variable appropriately.  It uses the same syntax
	as /set mangle_inbound and friends.

*** News 09/28/98 -- You may now strip ALL_OFF in $stripcrap()
	Just specify ALL_OFF as one of the items in the first argument.

*** News 09/28/98 -- Support for bzip2
	Scripts, help files, and other files that are loaded by the client
	may now be in .bz2 format as long as 'bunzip2' is available somewhere
	in your PATH.

*** News 09/28/98 -- New built in expando, $''
	Yea yea, i know i havent documented anything in a long time, so
	here is a bunch all together.  This new expando is just like $"",
	except that it only waits for the user to press one key before it
	returns.   BTW, $"" and $'' now use /input and /input_char as their
	underlying infrastructure, but put a synchronous wrapper on top of
	it, so it is now safe to mix the two input types (such as it is.)

*** News 08/03/98 -- New /on, /ON SWITCH_CHANNELS
	This is now hooked whenever any of your windows has its current
	channel set.  This includes the transitions from channel 0 to
	a new channel (when you /join) and from an old channel to channel 0
	(when you /part), and when you do a SWITCH_CHANNELS keybinding.
	If one of the channels involved is channel 0, then "0" will be the
	channel specified.
		$0	- The window that the change is happening in
			  *** You should not assume that this window
			      is the current window! ***
		$1	- The old channel
		$2	- The new channel

*** News 07/13/98 -- New stuff for stripping crap out of text
	Due to the incessant requests, ive given in and written a general
	purpose tool for stripping just about any kind of crap out of text.
	The tool is used in two ways:  Two new /set's that control just
	what (if anything) the client will strip out of incoming and 
	outbound lines to the *server*, and a new built in function that
	allows you to scour arbitrary text in arbitrary ways.  Lets get
	on to the gory details:

	This is the syntax for describing to the client just how to mangle:
		ALL	- Strip all of the below, except ESCAPE
		ANSI	- Convert all ANSI to logical characters, and
			  remove all "unsafe" ansi codes.  This does just
			  what $stripansicodes() does.  Note that to remove
			  any of the attributes below, you need to also
			  specify ANSI so that any ansi-sequence will be 
			  converted to their logical characters to get 
			  stripped.  If you dont specify ANSI, all ansi 
			  codes will be left untouched!
		BOLD    - Remove all bold_toggle chars.
		COLOR   - Remove all ^C color sequences.
		REVERSE - Remove all reverse_toggle chars.
		UNDERLINE - Remove all underline_toggle chars.
		BLINK   - Remove all blink_toggle (flashing) chars
		ROM_CHAR - Remove all rom_char characters
		ND_SPACE - Remove all non-descructive spaces
		ESCAPE   - This is a special one.  If you specify ESCAPE,
			   then any ESC characters will be converted to
			   <REVERSE>[<REVERSE>.  This will defeat your
			   ability to do ANSI, as there wont actually be
			   any ansi sequences left! ;-)

	You may specify any number of these modifiers by seperating them
	with commas.  For example, "BLINK,UNDERLINE"  means to remove both
	blink and underline characters.  You can negate the sense of an
	attribute by prepending it with a hypen.  So "ALL,-REVERSE" removes
	all the above attributes EXCEPT that it wont remove reverse chars.

	And now how to use this:
		/set mangle_inbound <description>
	This will mangle all inbound SERVER text in accordance with the
	description rules (as specified above).

		/set mangle_outbound <description>
	This will mangle all outbound SERVER text in accordance with the
	description rules (as specified above).

		$stripcrap("<description>" text)
	This will mangle the given text in accordance with the specified
	description (as specified above).

	*** Note that in general, using this feature is *very* expensive.

*** News 06/15/98 -- /TOPIC command extended to now unset topics
	You can now unset the channels topic by prefixing the channel
	name with a hyphen *and not specifying any arguments*.  That is
	to say, '/topic -#umbrella'   will unset the topic (if any) from 
	the channel #umbrella.  You can do '/topic -'  to unset the topic
	from the current window's channel.

*** News 06/10/98 -- EPIC now under the BSD license
	As ircII-current (post 4.4) has been changed to the BSD style
	license and ive been informed that permission has been given to
	retroactively offer the license to previous versions (eg, 2.8.2),
	I'm very happy to release EPIC under the BSD style license.
	See doc/Copyright for the specifics.

*** News 05/28/98 -- /on connect now passes in three arguments
	/on connect now offers the server's idea of what its name is
	as $2.  This may very well differ from what we think its name
	is supposed to be.

*** News 05/29/98 -- New built in functions $iget[r]matches() (Crackbaby)
	This works just like $getmatches(), but it returns the index number
	(in alphabetic order) of the array members rather than the item 
	numbers (in the order they were put into the array.)  The first
	argument is the array and the second argument is the pattern.
	$igetrmatches() is to $getrmatches() what $igetmatches() is to
	$getmatches().  Got it?

*** News 05/29/98 -- New /on, /ON SEND_CTCP
	When you send a CTCP request or a CTCP reply, this on will be
	hooked.  The arguments are:
		$0	- "PRIVMSG" for requests, "NOTICE" for replies
		$1	- Who the CTCP is being sent to
		$2	- The type of CTCP being sent
		$3-	- (Optional) Argument(s) to the CTCP being sent

*** News 05/29/98 -- New built in function $servnum()
	You give it a server name, it returns a server refnum for that
	server.  Returns -1 if the server isnt valid.

*** News 05/29/98 -- New type to /xdebug, "ALL"
	If you specify "ALL" to /xdebug, it will turn on all the xdebug
	options (as if you had specified the -x command line argument).
	Using "-ALL" will turn off all xdebug options.

*** News 05/29/98 -- New flag to /who, -u-i
	This flag will direct the client to add the 'x' argument to the
	WHO protocol command, which is used by the undernet ircd servers
	to give you invisible users.  (They changed it so that operators only
	see visible users now unless they use this special flag, so that the
	server can whine to everyone else about how youre abusing your power.)
	Don't use this server in any other situation, as itll probably 
	confuse the server or something.

*** News 05/29/98 -- $isconnected() no longer requires an argument
	If you dont specify an argument, 'from_server' is the default

*** News 05/29/98 -- /on dcc_connect compatability with BitchX
	/on dcc_connect will hook with $1 being "SEND" first, and if you
	dont grab that, it will offer the same exact thing with $1 being
	"GET".  This is for compatability with bitchx.

*** News 04/24/98 -- New built in command line argument (Texaco)
	The -s command line flag is now honored, and it instructs the
	client to not attempt to connect to a server on startup, but
	instead show you your server list (so you can /server to one)

*** News 04/24/98 -- New built in function $gettmatch() (Colten)
	Im not exactly sure what it does.  Someone tell me please? =)

*** News 04/24/98 -- New noise operator for /on, /on ?type
	This is kind of hard to explain, but here goes.  You may now specify
	an /on as having noisy type '?' (a question mark).  If you use this
	type, the /on will be executed silently (as if you had used /on ^type)
	but whether or not the /on actually supresses the default action will
	be determined by you at runtime.  You are expected to /return a value
	from the body of the /on.  If you return 1, the default action *will*
	be suppressed.  If you return 0, the default action *will not* be
	suppressed.

*** News 04/18/98 -- **hack** /on set "set-error *" is hooked by /set
	Whenever the user does something that would cause the /set command
	to emit an error, such as /set booya, where booya is not a valid 
	built in /set, you can hook /on set "set-error *" to suppress or
	redefine the error message.

*** News 04/18/98 -- /bind'ings also follow your /package settings
	So if you /unload package, then it will unload any /bind's that
	were set in that package.

*** News 04/18/98 -- /on 433, /on 437, /on 438 now pass server as $0
	These are the "nick collision" numerics, and now the server that
	rejected you is passed as $0.  I have cleared this change with Pana,
	so BitchX will do this as well shortly.

*** News 04/18/98 -- New built in function $getsets()
	If given an argument, it will return all the built in /set values
	that are matched by the argument(s).  If no arguments are given,
	it will return all built in /set values.

*** News 04/18/98 -- Client automatically sends DCC REJECT when you exit
	The client will automatically send CTCP DCC REJECTs on your behalf
	whenever you close or abort a DCC conenction.  This also will happen
	when you /exit.  This is so the remote peer will know that you will
	not be getting the transfer, and if theyre an EPIC or BitchX client,
	they will automatically close it for their end as well.

*** News 04/18/98 -- New /set, /SET SCROLLBACK_RATIO
	You may now specify, in percentage points, how much of the screen you
	want to scroll by every time you scroll forward or backward.  This
	value should be between 10 and 100, as the client silently truncates
	it to those extremes if you dont.

*** News 04/18/98 -- /DCC GET nick * now works
	You may specify a star as as special indication that you want to get
	all of the files that have been offered by "nick".

*** News 04/18/98 -- New set, /SET DCC_AUTO_TIMEOUT  (autoclose old dcc reqs)
	If at least N number of seconds has passed (where N is the value of 
	this set) since you offered or were offered a DCC handshake, 
	without the connection being accepted by either side, the client 
	will automatically close the offer on your behalf.

*** News 04/18/98 -- Doing /DCC SEND nick file   wont emit error second time
	Rather than emitting an unhelpful error message the second time you
	try to /DCC SEND to someone (perhaps because the first time you did
	the send the person wasnt on), duplicate /DCC SEND requests will now
	send a "booster" message identical to the original.
	(This works for /dcc chat nick as well)

*** News 04/18/98 -- /DCC RENAME can now rename chat connections
	To rename a chat connection, specify the original nickname as '-chat'.
	Then specify the old nick and the new nick.  Something like this:
		/dcc rename -chat oldnick newnick

*** News 04/18/98 -- DCC sliding window (/set dcc_sliding_window) modified
	You may now set this to any value, no matter how abusrdly large, and
	the client will keep up to this many packets outstanding, but never
	so many that the client will block.  Previously, if you set this value
	too high, the client would block.  Now it is safe to set this to 
	10,000, and acheive pretty good throughput.

*** News 04/18/98 -- New /dcc command, /dcc closeall	
	This closes all of your registered DCC connections, leaving your DCC
	list very empty.

*** News 04/18/98 -- Windows may now have zero visible lines
	You may now do /window size 0 to make a window that is nothing but
	a status bar.

*** News 04/18/98 -- New /set, /SET ND_SPACE_MAX
	For those who want to cap the number of non-destructive spaces that
	may appear in any one line of output, you may set this variable to
	control the number the client will parse before ignoring them.  The
	idea is simliar to /set tab_max

*** News 04/18/98 -- New flag to /xecho
	If you use the double-hyphen flag to /xecho, it will stop all
	argument parsing right there and take the rest of the arguments as
	the text to be echoed.  This is useful if the first word you want to
	output starts with a hyphen
			/xecho -- -s booya booya booya boo
	outputs 
			-s booys booys booys boo

*** News 04/18/98 -- New flags to /lastlog
	There is a new flag to /lastlog, -all which specifies that the given
	lastlog level to search for is to be all qualifying levels.  You can 
	turn off a given level by using a double hyphen (ala --msgs).  So, if
	you want to see everything in your lastlog EXCEPT your msgs, do this:
			/lastlog -all --msgs
	Another new flag is -reverse.  The -reverse flag instructs the client
	to start from the END of your lastlog and go BACKWARDS through the 
	lastlog when outputting.

*** News 04/02/98 -- New command, /userip
	This is the front end to the Undernet command by the same name.  It
	is exactly like /userhost except it returns the ip address instead of
	the named userhost.

*** News 04/02/98 -- New command, /usleep 
	The /usleep command pauses the client for some number of microseconds.
	specified as the argument to the command.  This command will probably
	be rolled into /sleep at some future time, Watch for this.

*** News 04/02/98 -- $cparse() changed to support bitchx-syntax
	If the first argument passed to $cparse() is an extended word, then
	only the first argument will be used for the output, and any numeric
	expandoes in the first argument will be expanded with the rest of the
	arguments to $cparse() -- ala:
		$cparse("This is a $0 test" cparse)

*** News 04/02/98 -- /window create now works properly for GNU screen
	Kanan did the work for this, you can thank him. ;-)

*** News 04/02/98 -- You may now /msg a DCC RAW connection.
	To send a message to a DCC RAW connection, simply msg to =NUM
	where NUM is the number that was previously returned by $connect().
	Obviously, you can also /query a DCC RAW connection with this, making
	it usable for protocol connections to remote servers.  I have fixed
	and returned the 'mudirc' script, which demonstrates how to make epic
	act as a relatively useful mud/moo client.

*** News 03/24/98 -- Scope resolution operator for variables
	The client now supports a way to differentiate between local 
	variables and global variables having the same name.  As always, 
	an expando with no modification will resolve to a local variable 
	if it can, and a global variable it cannot.  However, if the 
	variable name is prefixed with a single colon (either as an lval 
	OR as an rval), then the variable will be expanded *only* as a 
	local variable, and if the local variable does not exist, it will 
	expand to the false value.  If the variable name is prefixed with 
	two colons (again, either as an lval or an rval), the variable will 
	be expanded *only* as a global variable, even if a local variable 
	by that name exists.  Here are some examples:

			assign foo 1
			local foo 2
			@ ::foo = 4		# Sets the GLOBAL variable
			@ :foo = 6		# Sets the LOCAL variable
			@ foo = 8		# Sets the LOCAL variable
			@ bar = 12		# Sets the GLOBAL variable
			$::foo		# is '4'
			$:bar		# is the false value
			$bar		# is 12
			$foo		# is 8

*** News 03/24/98 -- New /on, /ON UNLOAD
	As a courtesy to script writers that use /package in order to keep
	track of all of the aliases and so forth that they create in their
	scripts, /ON UNLOAD will be hooked with the package name (as supplied
	to /package) when the user does /unload <packagename>.  Presumably, 
	you could use this to close open logfiles or $connect()s or what
	ever garbage collection that is needed.

*** News 03/24/98 -- Built in function $tdiff() enhanced (as a demo)
	To demonstrate the proof of concept, the $tdiff() function has been
	expanded to support decimals.  For example:
		$tdiff(60.01) returns "1 minute 0.01 seconds"
	If you like this support and want to see it applied to the other
	$tdiff*() functions, let me know.

*** News 03/24/98 -- New /set, /SET MIRC_BROKEN_DCC_RESUME
	Unfortunately, due to popular demand, i have broken my previous
	promise to never #define MIRC_BROKEN_DCC_RESUME, a feature which
	breaks both the spirit and the letter of RFC 1459.  As a comprimise,
	the ability to use this feature is controlled through a new set by
	the same name, /SET MIRC_BROKEN_DCC_RESUME which will now and forever
	default to OFF.  (Since /set's can be changed by a script, there will
	not ever be a need for me to change this default to ON.)  I stand by
	my previous statement that any script that uses this feature is 
	broken and you should pester the script author to take it out.

*** News 03/24/98 -- /ON DCC_OFFER extended
	If you offer a dcc send to another user, then /on dcc_offer will
	hook with additional arguments:
			$2   - filename being offered
			$3   - size of the file being offered

*** News 03/24/98 -- Built in function $isnumber() changed
	You may now specify a base argument, prepended with a 'b'.
	For example, to determine if '014' is a number in base 6,
		$isnumber(014 b6)	or	$isnumber(b6 014)
	It makes no difference whether the base is the first or the
	second argument:  but you must put a 'b' before it either way.
	Bases 0 through 36 are supported.  Base 0 (the default) is a 
	special base that attempts to auto-grok the argument for its
	base (see the man page on strtol(3))

*** News 03/24/98 -- New built in function $rest(start text) (Matt)
	This function works like $restw(), but it works on strings.  It
	returns 'text' beginning at the 'start'th character

*** News 03/24/98 -- /bind metaX now shows all of the meta-X bindings
	If you do something, like say, /bind meta1, then the client will
	show you all the bindings you have that start with meta1.

*** News 03/24/98 -- The slash (/) always recognized as a command char
	As a consideration to scripters, the client will always accept
	the slash as a command character, even if /set command_chars does
	not include it.  You may then safely and with a free conscience 
	do //built_in_command without having to lie awake at night worrying
	that someone would take away the slash from you.

*** News 03/06/98 -- Add or delete specific levels from /window level
	You may now add or delete levels from your existing /window level
	by prepending the level(s) with a plus ('+') for addition and a
	minus ('-') for subtraction.  For example:

		/window level all	-- sets the window level
		/window level -crap	-- removes JUST crap level
		/window level +crap	-- adds crap level back.

*** News 03/04/98 -- You may now use "special chars" in your input prompt
	Previously, you could use ansi sequences to turn on bold and so
	forth in your input prompt.  Now you may use ircII's "special chars"
	to turn them on as well (like ^B, ^C, etc).  The flipside of this is
	that you may only use "safe" ansi codes, if you use them at all, since
	unsafe ansi codes will be stripped out by the client.  This is 
	intended to be a feature, not a bug.

*** News 03/04/98 -- New character for $info(o)
	If the return value from $info(o) includes the character 'm', then
	you know that you should /set color instead of /set control_color_c.
	This can be used in conjunction with the next note...

*** News 03/04/98 -- IMPORTANT!  /SET CONTROL_C_COLOR RENAMED TO /SET COLOR
	It is important to note this change.  If you have /set control_c_color
	in your script, you should do one of two things to work around this:

	1) Use a compatability /set for older epic4pre1.* clients:
		/on ^set "COLOR *" set control_c_color $1

	2) OR use a compatability /set for newer epic4pre1.* clients:
		/on ^set "CONTROL_C_COLOR *" set color $1

*** News 03/03/98 -- Correct support for "nondestructive spaces".
	You may now use the ^S special char as a "non destructive space".
	This is emitted by the ansi parser when you do the ^[[<N>C code. 
	The general idea of a nd space is that any color/highlight/etc that
	is active *will not* be output for the space (it is left totaly 
	blank of all attributes)

*** News 03/02/98 -- New flag to /input, /input -noecho
	If you want to input a password, or some other sensitive information,
	you may use the -noecho flag to /input to suppress the echoing of
	the characters as they are typed.

*** News 03/02/98 -- Ten new /set status_user(10-19) supported.
	You may now do /set status_user10 through /set status_user19.
	They will show up on your status bar via the %{1}0 through %{1}9
	expandoes.  See below for explanation of how %{N}M works.

*** News 03/01/98 -- Uppercase user modes supported
	Now all 26 of the uppercase user modes are supported, along with
	the 26 lower case modes.  Should cause the client not to crash
	on DALNet and so forth.

*** News 02/27/98 -- Semantic changes for status line expandoes.
	It is now possible (though there is no support for it yet), to
	specify a "map" for your status line expandos.  The general syntax
	is:	%{N}M     where 'N' is some number, and 'M' is the expando
	that you want.  Note that for different values of 'N', you will get
	different values for 'M'.  The "default" map is 0.  To give you an
	idea how it works, try using %{0}S and see that it is the same as
	%S.  I have included (for testing purposes) a single expando %T on
	map #1.  Use %{1}T and see that it is *different* than %T is.  It
	is expected that this will allow an effectively unlimited number of
	status line expandos in the future.

*** News 02/27/98 -- New built in function, $isnumber(text)
	You pass it a number, it tells you if it is a number or not.  It
	correctly handles negative numbers, and hexadecimal/octal.  More or
	less all it does is call strtol(input, NULL, 0).  See the man page 
	for strtol to see how that works.

*** News 02/27/98 -- Four new mask types for $mask().  
	There are now four new types of masks for $mask():

		N		Z is a hostname		Z is an ip address
	-------------------------------------------------------------------
		10		*!*@h.d			*!*@d.*
		11		*!*u@h.d		*!*u@d.*
		12		n!*@h.d			n!*@d.*
		13		n!*u@h.d		n!*u@d.*

	With the following modifications:  In the ``local'' portion of
	the hostname, all sequence of numbers are substituted with a
	single '*'.  That is to say, 
		$mask(10 nick!user@ppp-147-0-52-129.frobitz.com)
	returns
		*!*@ppp-*-*-*-*.frobitz.com


*** News 02/22/98 -- New built in function, $stripansicodes(text)
	This function will "normalize" the text, suitable for displaying.
	Generally, "normalization" has two aspects.  The first aspect is that
	if DISPLAY_ANSI is ON, then all ansi codes are either converted into 
	logical codes (^C sequences, or ^B, or ^_, etc), or they are stripped
	out entirely.  If DISPLAY_ANSI is OFF, then just the escape is mangled,
	and the rest is left intact.  ^C codes are also converted to a form
	that correctly manages bold and blink attributes.  THIS IS THE SAME
	CODE THAT PREPROCESSES YOUR OUTPUT FOR THE SCREEN.  So you can be
	assured that What You Get Is What Youre Going To See.

*** News 02/22/98 -- New built in function, $printlen(text)
	This function returns the number of _printable characters_ in the
	text _as it would appear on the screen_.  The string you pass must
	not contain any ansi sequences, as they are interpreted as printable
	characters.  Use $stripansicodes() to remove them.

*** News 02/22/98 -- New math operator, **
	The ** operator now acts as an exponential operator.  All it does is
	call pow(), so you can specify whatever arguments you want with it.
	Yes, it correctly handles negative exponants and fractional exponants,
	as long as your math library does!

*** News 02/22/98 -- New built in function $remws(lhs / rhs)
	This function returns all of the words on the rhs of the the /,
	except that any words that appear on the lhs of the / have been
	removed first.

*** News 02/22/98 -- New built in function, $pad(len char string)
	This allows you to pad a string out to a certain length with
	a specified character.  If the string is too long, it is NOT
	truncated (as opposed to $[x]y).

*** News 02/18/98 -- Twenty new meta maps
	Yep.  Now there are thirty nine of them.  META1-META39.  
	META4 is still, as always, sticky.

*** News 02/18/98 -- New built in function, $channel()
	This returns some information about who is on your channel.  It
	returns a sorted list of nicknames prepended by two characters. 
	The first character is a '@' if the person is a channel operator, 
	and '.' if they are not.  The second character is '+' if the person
	is a channel voice, and '.' if they are not, and '?' if we're not
	sure either way.

*** News 02/16/98 -- New built in function, $iscurchan()
	This takes a channel argument, and will return '1' if the specified
	channel is the current channel for *any* window that is bound to the
	current server.  It returns 0 otherwise.  This is intended to allow
	for more flexibility for testing for current channels in /ON's.  EG:

	Rather than:
		/on ^send_public * {if ([$0] == C) {...} {...}}
	Try doing:
		/on ^send_public * {if (iscurchan($0)) {....} {...}}

	This will allow you to do the "right thing" even when $0 is not the
	current channel for the current window. 

*** News 02/14/98 -- New built in functions: $getuid(), $getgid(), $getpgrp()
	As well as $getlogin().  These correspond to the system calls by the
	same name.  Do a lookup on the man page for information on what they
	return.

*** News 02/14/98 -- Changes to how ansi codes are handled
	IF YOU HAVE /SET DISPLAY_ANSI ON:
	New code contributed by fireclown now eradicates all ansi codes
	from the output before it is counted and displayed.  Any codes that
	are recongized and are "safe" (such as ^[[<X>m, where <X> is one of
	the color codes, or bold/blink/reverse/underline), are converted to
	the corresponding logical character.  All other codes are blown away
	entirely.  Not just the escape, but the whole thing (at least as much
	as we can recognize it.)

	IF YOU HAVE /SET DISPLAY_ANSI OFF:
	Just like before, the escape is mangled into reverse, and the rest
	of the string is untouched.

	Also, since all ansi color codes are converted to ^C codes, then
	if you /set control_c_color OFF, all color will be inhibited.  
	Conversely, if you /set control_c_color ON, all color will be 
	honored.  IT NO LONGER IS RELEVANT WHAT YOUR /SET DISPLAY_ANSI
	VALUE IS TO DETERMINE HOW YOU SEE COLORS!

*** News 02/14/98 -- New ^C codes
	The Control-C Color codes from 00 to 15 are now in "mirc" order
	(which is also used by BitchX).  This has been changed to make it
	easier on scripts that want to be BitchX/EPIC compatable.  The 
	ANSI order codes have been moved to:
			30 - 37		Ansi Foreground colors
			40 - 47		Ansi Background colors
			50 - 57		Emphasized fg/bg colors
	Note that 30-37,40-47 corresponds directly to the ^[[<X>m codes
	and that 50-57 is either the BOLDED colors of 30-37, or the BLINKING
	colors from 40-47, depending on whether or not it is on the lhs or
	the rhs of the color code.

*** News 02/14/98 -- Two new debugging flags to /who
	/WHO -d  (diagnose) will show the contents of the who queue for the
		server for your current window
	/WHO -f  (flush)    will remove all the entries of the who queue for
		the server for your current window.  You must *NEVER* do this
		while a WHO request is still pending, or the client will 
		fall over.  If you report this as a bug, i will laugh at you.

*** News 02/14/98 -- New built in function, $rigtype()
	You give it some number of ignore levels, it will give you all of the
	patterns (exact patterns) that match *only* those levels (no more,
	no less.)  Dont ask, this is what was requested. ;-)

*** News 02/14/98 -- New built in function, $igtype()
	You give it one of your ignore patterns (must be EXACTLY the same.
	This is not for use as a matcher.  Use $rigmask() for that) and it
	will give you the IGNOREd levels on that pattern.

*** News 02/14/98 -- New built in function, $winlevel()
	You give it a window description (name or refnum) and it gives
	you the lastlog levels that are bound to that window.  Simple,
	aint it?

*** News 02/14/98 -- New /set, /SET HIGH_BIT_ESCAPE (0 | 1 | 2)
	This controls the behavior of how the client handles 8 bit characters.
	If this is set to 0, the eighth bit, if any, is always stripped to
	yeild a 7-bit ascii value.  If this is set to 1, then if the 'km'
	termcap variable is set (hint: it almost always is), then the eight
	bit character <meta>-X will be converted into a <ESC>-X sequence.
	If this is set to 2, then the client will leave the character alone
	and parse it as a normal 8 bit character.

*** News 02/14/98 -- New /on, /ON SERVER_LOST
	This is (hopefully) hooked any time youre disconnected from any
	server for any reason.

*** News 02/14/98 -- Variable modifier : (a colon) implicitly declares local
	You may now *** as an lvalue *** prepend the variable name with
	a colon to impliclity declare it as a local variable.  You must not
	use the colon in any rvalue context, or in the /fe or /foreach 
	commands, as the variable is already declared local there.  You need
	only use the colon once.  Any future references will automatically 
	use the local variable because it will have been declared that way.

*** News 02/14/98 -- New flag to /xecho, /xecho -s
	/xecho -s attempts to emulate the internal "say()" function, and
	promises not to output anything if you have run the command with the
	^ modifier to suppress any output.  (Normally, all /xecho's always
	are output, even if you did ^cmd to supress output)

*** News 02/14/98 -- New status line expando, %. (percent-dot)
	This outputs the current contents of whatever was set with the
	/window status_special command (see below)

*** News 02/14/98 -- New /window command, /WINDOW STATUS_SPEICAL
	This /window command sets some window-specific information that
	you would like to see displayed on your status bar.  It will be
	displayed in the status bar using the %. expando (thats a dot)

*** News 02/14/98 -- New built in function, $chop(num text)
	This function returns "text" with the LAST "num" characters
	removed.  If "num" is negative, nothing happens.

*** News 02/14/98 -- /set status_user<3-9> changed
	They now go to your current window rather than to a NON current
	window.  I did this because it seemed everyone wanted it that way.

*** News 02/14/98 -- New set, /set current_window_level <LEVELs>
	This set will allow you to specify some number of levels as always
	going to the current window *if and only if* the current window is
	attached to the same server as the output to that level.  Otherwise,
	it would go to the "normal" window.  This is useful for people who
	want informational stuff to always go to a visible window, so they
	may choose to do  /set current_window_level CRAP

*** News 02/14/98 -- $mychannels() can take an argument
	If the argument is a number, then it is taken as a server refnum
	that youre interested in.  If it is not a number, then it is taken
	as a window name that youre interested in.  But wait!  There's more.
	If you must have a need to specify a window refnum instead of a
	name, you can prefix it with the '#' character and thatll work too.

*** News 02/14/98 -- Documentation updated
	Youll notice this file is way behind the work ive done.  So today
	im gong to go through and update all the stuff ive done since the
	last time i documented it (2 and a half months).  They all have
	todays date, but theyve been done at various times recently.

*** News 12/29/97 -- New built in function $ischanvoice(nick #channel)
	This function returns one of three values depending on the current
	+v status of the given nick on the given channel.
		 0	-- The person is not +v
		 1	-- The person is +v
		-1	-- I dont know

*** News 12/19/97 -- New built in function $mask(type address)
	Returns a pattern suitable for /ignore'ing or banning.  If 'type'
	is N, and the address is Z (either n!u@h.d, or n!u@d.h, where the
	former is a host.domain and the latter is an ip address), then the 
	return value is:

		N		Z is a hostname		Z is an ip address
	-------------------------------------------------------------------
		0		*!u@h.d			*!u@d.h
		1		*!*u@h.d		*!*u@d.h
		2		*!*@h.d			*!*@d.h
		3		*!*u@*.d		*!*u@d.*
		4		*!*@*.d			*!*@d.*
		5		n!u@h.d			n!u@d.h
		6		n!*u@h.d		n!*u@d.h
		7		n!*@h.d			n!*@d.h
		8		n!*u@*.d		n!*u@d.*
		9		n!*@*.d			n!*@d.*

*** News 12/19/97 -- Changes to $winchan() and $chanwin(), now synonyms
	Since there is no compelling reason to make a distinction between
	$winchan() and $chanwin(), and since the specific distinction that
	was made caused no end of problems for both me and scripters, the
	two troubled functions are now synonyms of each other.  If you pass
	it a channel, it will return a window refnum.  If you pass in anything
	that isnt a channel, it will be considered a window refnum or name,
	and the current channel of that window will be returned.

*** News 12/19/97 -- New built in function $winrefs()
	This returns the refnums of every currently open window, both hidden
	and invisible.  While the order of the refnums is *EXPLICITLY*
	undefined, it follows in traverse_all_windows() order, which usually
	returns windows in physical order on their screens, and the invisibile
	windows are last.  You may NOT assume this, though. ;-)

*** News 12/19/97 -- New built in function $querywin(target)
	Given a query target, this returns a window that has that target
	as its current query.  If more than one window has the specified
	target as its query, which window is returned is indeterminite.

*** News 12/19/97 -- New built in function $uname(description string)
	This returns the entire argument string as-is except with the 
	following modifications, as retreived from the uname(2) system call.

	Any instance of		will be replaced with
	---------------------------------------------------------------------
	%m			The "machine" (architecture)
	%n			The "Node" (hostname)
	%r			The "Release" of the operating system
	%s			The "Name" of the operating system
	%v			The "Version" of the operating system
	%a			The logical value of "%s %n %r %v %m"
	%%			A single % 

*** News 12/19/97 -- New built in pattern matcher
	A new pattern matcher has been written.  It is expected to be
	bug-for-bug compatable with the old pattern matcher.  If you see
	any discrepencies, no matter how obscure, please let me know.
	The reason for the new matcher is that it uses an iterative algorithm
	rather than a recursive algorithm, and so is expected to be faster.

*** News 12/19/97 -- New built in commands, /PACKAGE and /UNLOAD
	The /PACKAGE command allows you to "group" all of the assigns,
	aliases, and ons (and stubs) in a script under a common package
	name.  When you use the /package command, any alias/assigns/on that
	are registered until the end of the file are given the current
	package name.  You can change the package name anywhere in a script,
	even if you have previously used /package.  However, packages only
	last until the end of the file -- they never propogate upwards.
	Scripts that are loaded by other scripts may put themselves under 
	a different package name than the one being used by the loading
	script.  By default, a script is put into the package that is
	currently active in the script that loaded it (or * if no package
	is active)

	The idea of a logical grouping is useful for the purposes of 
	removing the entire package later.  So the /unload command allows
	you to remove a package that was constructed as above.  

	The idea behind all this is to allow a script, and anything that
	it loads, to be easily removed whenever you want.  By simply putting
	a /package <something> command at the top of a script, then you can
	later do /unload <something> and everything that was loaded in that
	original script will be removed.

	As these commands will surely be used in the future in ways that
	are not really easy to describe now, you should be careful to not
	pass more than one argument to /package or /unload.  Even though
	multiple arguments may be recognized now, for forwards compatability,
	you should not get in the habit of doing that.

*** News 12/19/97 -- New built in function $cparse(text)
	The $cparse() function returns the given text without modifications
	except for the following exceptions:

	Any instance of		will be replaced with ^C codes to change the 
				text to		text to		background to
	---------------------------------------------------------------------
	%k	%K	%0	black	  	bold black	black
	%r	%R	%1	red  	  	bold red	red
	%g	%G	%2	green	  	bold green	green
	%y	%Y	%3	yellow 		bold yellow	yellow
	%b	%B	%4	blue		bold blue	blue
	%m	%M	%5	magenta		bold magenta	magenta
	%p	%P		magenta (think: purple)
	%c	%C	%6	cyan		bold cyan	cyan
	%w	%W	%7	white		bold white	white
	%F			Flashing attribute turned on
	%n			All colors turned off
	%N			Don't put a clear-color tag at the of output
	%%			A single %

	Of course, you need to have /set control_c_color on to see the 
	colors, so this is an easy way for a script to be color-aware 
	while not forcibly imposing colors on an unwilling user. ;-)

*** News 12/19/97 -- Built in functios $winnum() and $winnam() extended
	They can now take a window description (eg, a window name or a 
	window refnum) as an optional argument, and will return the info
	for that window, if it exists.  As always, with no arguments it
	still defaults to the current window.

*** News 12/18/97 -- New /window command, /WINDOW QUERY
	This allows you to set the query nicknames for any window that
	you can specify.  The /query command now uses it.

*** News 12/18/97 -- New attribute, Blinking
	This is a new character attribute, just like underline and bold
	and reverse.  Not all terminal emulations will support it, but
	the linux and freebsd consoles do.  By default, the blink key
	is ^F (think of flashing), and you can input that by /bind'ing 
	something to BLINK.  You can turn it off with /set blink_video off

*** News 12/16/97 -- New script: "man"
	After you do /load man, you will have available a "man" alias
	that will search all of the subdirectores if the man-style help
	pages for the topic youre looking for.  It handles multiple
	sections ok.  Try /man away   to get an idea.

*** News 12/16/97 -- New argument passed to /on join.
	A fourth argument is now passed to /on join.  Here is a summary of
	the the arguments to /on join.
		$0	Nickname of who joined
		$1	Channel that was joined
		$2	Userhost of who joined
		$3	Zero more of the strings (+o) or (+v) 
			    (only on avalon 2.9 servers)
			    -- (+o), if present, is always $3.
			    -- (+v) can be either $3 or $4.

*** News 12/16/97 -- New argument to /help, -WAIT
	If you do /help -wait, then it will not return until the help system
	has finished whatever its doing.  If the user is not currently doing
	/help, it will return immediately.

*** News 12/16/97 -- New argument to /xecho, -NOLOG
	The -NOLOG argument to /xecho inhibits the current output line from
	being logged, either at the /set log level or the /window log level.
	Im not sure why you would want to do this, but Kasi asked for it.

*** News 12/15/97 -- New built in function, $leftpc()
	This function returns the LONGEST string of "text" that contains
	"N" printable characters, where "N" is the first argument to the
	function and everything else is text.  For example:
		$leftpc(10 onetwothre$chr(27)[34mefour)
	return
		onetwothree<blue attribute>
	Note that any non-printable characters that appear after the N'th
	character *will* be included in the output.  This is intentional.

*** News 12/11/97 -- New built in function, $rigmask()
	This works just like $igmask, but in reverse.  You provide it
	a nick!user@host pattern, and it returns to you all the ignores
	that would be triggered by that pattern.

*** News 12/05/97 -- Ten new meta keybindings, META10 through META19
	You now have 19 meta keys to work with.  The client attempts as
	much as possible to conserve space, so this isnt a RAM buster.
	If you dont use the new meta maps, the only space reserved for them
	will be a NULL pointer. ;-)  Meta4 is still sticky.

*** News 12/03/97 -- New flood protection, change to /ON FLOOD
	On flood now takes one more argument:
		$0	The person doing the flood
		$1	The type of flood theyre doing
		$2	The channel (if any) theyre flooding on   (NEW!)
		$3-	The text of the flood.

*** News 12/02/97 -- New flags to /CLEAR and /UNCLEAR
	They are -HIDDEN which clears or unclears all the hidden windows,
	and -VISIBLE for all of the visible windows.  Theyre mutually
	exclusive.  The last one used prevails.  Use -ALL if you want all
	windows to be adjusted.

*** News 12/01/97 -- New built in command, /UNCLEAR
	This command works just like /CLEAR, except it nails the bottom of
	your scrollback display onto the bottom of the screen (or as low as
	possible if you dont have a full screen of display yet)

*** News 12/01/97 -- New built in function, $msar()
	This is a "multiply substituting" $sar().  You give it any number
	of delimited patterns, and it will substitute all of the patterns
	on the text.  Of course, the obvious limitation is that the text
	may not contain the delimiter character.

*** News 11/27/97 -- New built in function, $randread() (srfrog)
	Given a filename, this returns a random line from that file. 
	Yes, it does ~ completion...

*** News 11/25/97 -- New built in function, $igmask()
	You give it a pattern, it returns all of the ignore patterns that
	are matched by the pattern.

*** News 11/20/97 -- New flag to /timer, -WINDOW <windesc>
	This flag will allow you to specify that the timer should go off
	in the specified window.  Please note that the limiations about
	timers that go off in hidden windows still applies.  Use this flag
	with caution until the problem with timers in hidden windows is 
	addressed.

*** News 11/20/97 -- Changes to /dcc close
	You may now use the words "-all" or "*" to indicate that all of the
	dcc's of a given type or dcc's to a given nick should be closed.
	For example:
		To close all of the dcc's you have open to "foobar",
			/dcc close * foobar 
		To close all of the dcc chat's you have open to anyone,
			/dcc close chat *

*** News 11/19/97 -- New built in function, $count()
	This function returns the number of times (possibly overlapping) 
	that a given string appears in some text.  For example:
		$count(. one.two.three)		returns 2.
		$count(ll llll)			returns 3 (think about it).

*** News 11/12/97 -- New /ON, /ON DCC_OFFER
	This hook is activated whenever you make an OUTBOUND DCC OFFER.
	(Kind of the complement to /on dcc_request).  
	Presently, it has exactly two arguments: 
		$0 is the person to whom you are offering the dcc, 
		$1 is the type of dcc you are offering.
	It is possible that more args could be added in the future on request.

*** News 11/11/97 -- New /SET, /SET STATUS_TRUNCATE_RHS **** RECALLED *****
	This set determines which side of the %> in your status bar should be
	truncated if the status bar is wider than the screen.  If this is set
	to ON, then the right hand side of the %> will be truncated back to
	80 characters.  If this is set to OFF, then the left hand side of the
	%> will be truncated so as to allow everything to the right of %> to
	be kept.
		***** This feature is non-operative due to some *****
		*****      serious problems.  It will remain    *****
		***** non-operative until such time it is fixed *****

*** News 11/10/97 -- New script, "pipe"
	This script implements the $pipe() function, which demonstrates how
	you can use /on exec to capture the stdout of an exec'd command and
	return it from a function (essentially acting as a pipe or a backtick
	type operation).  It may not be universally useful as-is.  Feel free
	to tweak it to your taste.

*** News 11/07/97 -- Improvements to /ON
	As part of a larger re-organization of the /on command, three noticible
	changes to the semantics of /ON have been made:

	1) The @ and # "server-specific" qualifiers have been withdrawn.
	   If you used this feature, and desperately need to retain it, let
	   me know and ill work something out with you.  I did not reimplement
	   them on the understanding that nobody would miss them.
	2) Doing /on ^type ^"nick" now works as you would expect, that is to 
	   say, it really does nothing, and it would suppress the default
	   behavior.  This is a lower-cost way to do  /on ^type "nick" #.
	3) You may now do /ON SEND ("send" is just an exmaple, this works for
	   any hook) to see all of the hooks for types that begin with "send".
	   Previously, it was impossible to get a listing of more than one
	   hook type at a time.

*** News 11/05/97 -- Improvements to /TIMER
	/TIMER now attempts to keep track of what the current server and
	current window were when the timer was registered, and will revert
	back to that window/server when the timer expires.  This should prove
	to be much more intuitive.

*** News 10/28/97 -- New flag to /xecho, /XECHO -R
	This flag dumps a raw string to your terminal emulator.  It is
	imperative to notice that the client does *no interpretation* of the
	string.  It is dumped right to your terminal by the /xecho command,
	so it never gets backed by your window/screen.  That means you should
	really only use this to output strings that act as a state-change 
	command to your terminal (eg, setting the xterm title bar, or doing
	an unflash.)  If you output "modifying" codes (like setting colors or
	attributes), it wont get to your window display and will be lost on
	the next window redraw.

*** News 10/27/97 -- New /set, /SET NO_FAIL_DISCONNECT (ON|OFF)     (nuke)
	When this is set to ON, the client will NOT automatically terminate
	your connection to the server if a write() fails.  This is most useful
	for nuke and other people who have static ip's connected to routers
	that retain packets while you are disconnected.  The default is OFF,
	since for most people, a write() failure is a fatal condition.

*** News 10/26/97 -- New built in commands, "break", "continue", "return"
	These built in commands are used for controlling the flow of code
	execution.  They work 99% like they would in C.  When youre in a 
	"while" or "for" loop, the "continue" command will immediately
	terminate the current loop and go back to the top (the 'for' command
	will run the 'iteration' part as it does in C).  When you do a 
	"break" command in "while" or "for", it immediately terminates the
	loop entirely and moves to the next command after the "while" or "for".

	The "return" command is used to immediately terminate execution of
	an alias or on-event.  If an argument is specified, it is assigned to
	the local variable FUNCTION_RETURN.  Those of you who use the old
	"return" alias that just assigns to function_return will find that it
	still works, but it wont terminate the execution of the current alias
	or on (eg, you wont notice any changes.)  Youre encouraged to take 
	advantage of the new functionality. =)

*** News 10/26/97 -- "FUNCTION_RETURN" now a normal local variable.
	When you make a function call, the "function_return" variable is now
	instantiated as a regular local variable.  Previously, the
	function_return variable was trapped for special treatment and was
	a write-only value.  Now it is a read/write variable, so you can use
	it just like any other variable.  The one exception to this is that
	you may assign to FUNCTION_RETURN in aliases that are called from the
	function that instantiated them (as opposed to other local variables
	which are not honored in called aliases.)  This was done out of
	neccesity to be compatable with scripts that presume that you may
	assign to FUNCTION_RETURN and it will carry to the closest enclosing
	function call.  Since FUNCTION_RETURN is stored on the calling stack,
	the arbitrary limit of 128 nested function calls no longer exists.

*** News 10/25/97 -- New Status bar expando -- %K
	This expando will be displayed on the status bar if you are currently
	in scrollback mode in that window.  Should be good for people who
	forget. ;-)  What it actually displays is controlled by a set, 
	/set status_scrollback.  The default is "(Scroll)"

*** News 10/20/97 -- Greatly reduced CPU usage in pre0.30
	Some profiling work showed some really monsterous CPU hogs.  With
	a few minor, yet strategic changes, CPU usage for the testers has 
	been reduced from 60% to 90% compared to pre0.28. (ive seen cpu usage
	under 2:00 a day for someone on 10 channels running with all usermodes
	active running clonebot-detection scripts.)

*** News 10/20/97 -- New built in functions, $regcomp(), $regmatch(), etc.
	There are four new built in functions that give you an interface
	to the posix.2 "regex" pattern matching functionality:

	$regcomp(<pattern>)
		given a regex pattern, it will return a printable string
		representation of the compiled pattern.  You must later pass
		the return value of this function to $regfree() to avoid 
		memory leaks.  The pattern is assumed to be case insensitive,
		and cannot make use of any \(...\) subexpressions.

	$regexec(<compiled> <string>)
		given a compiled regular expression and some text, it will
		return 0 if the string is matched by the compiled pattern,
		or will return nonzero on error.  The return code can be
		passed to $regerror() for specific information.  Of note, 
		the return value '1' means the pattern didnt match.

	$regerror(<error code> <compiled>)
		given an error code and the compiled regular expression from
		which it was generated, this gives you a human readable error
		message.  If the error was from $regcomp(), then you can pass
		the error code "-1", which will then be subtituted with the
		last $regcomp() error code.  Otherwise, error code should be
		the return value from $regexec().

	$regfree(<compiled>)
		given a compiled regular expression, this returns any resources
		that were used by the regular expression.  If you fail to 
		$regfree() everything you $regcomp(), then you will expose a
		memory leak, and that wont be my fault. ;-)  You must not use
		a compiled regular expression after you have called $regfree()
		on it.  To do so will result in chaos.

			
*** News 10/20/97 -- New command line flag, -B
	This does what the stock client's -b flag does.  It loads your .ircrc
	file right at client startup time rather than waiting for the client
	to connect to a server.

*** News 10/20/97 -- New math operators << and >>.
	These are the left-bitshift and right-bitshift operators.  Currently
	they are only available in their binary forms.  There is no support
	for <<= and >>=, as support for that would be problematic.  Since you
	can just do the longhand assignment, their ommision is not a big deal.

*** News 10/16/97 -- New built in functions, $fnexist(), $cexist()
	These two functions return 1 if the given built in function, or
	the given built in command, exist.  It returns 0 if they do not
	exist.  The idea use for these is to provide an alias or function
	that emulates the built in function/command youre looking for, if
	theyre absent (as in older versions).  Of course, this is more of
	a forward looking feature, as it wont help you on older releases. =)

*** News 10/15/97 -- New /set, /SET STATUS_DOES_EXPANDOS (ON|OFF)
	This /set controls whether or not any expandos (eg, $vars or $func()s)
	in the status line should be expanded or left alone.  Please note that
	this option is *very expensive*, and so you should not turn it on
	unless you really do have expandoes in the status_format, and you 
	really shouldnt put expandoes in the status_format if you can possibly
	get away with it. ;-)

*** News 10/15/97 -- New built in function, $currchans(server)
	If you specify an argument, this function will return quoted words
	containing a server refnum and a channel name.  The channels to be
	returned are channels that are currently considered to be "current
	channels" for some window that is connected to the specified server.
	If you specify a server of -1, then $lastserver() is assumed.  If you
	specify no arguments, then all current channels will be returned.
	It is expected that this will be useful to determine if a given
	channel on a given server is a current channel on *any* window. 
	The return value of this function uses double quotes, so it is
	suitable for use in /fe, or to pass to $rmatch().  Just remember to
	glob against "* #chan" (with the quotes!) for any server.

*** News 10/15/97 -- Change to /window (Colten)
	If you specify a window refnum where /WINDOW excepts to find a command,
	then that window will be used as the target window for any further
	commands, *but*, that window will not be made the current window (as
	it would if you had used REFNUM <refnum>).  This is useful for 
	changing an attribute for a specific window but you dont want the
	current window to change.  eg:

			/window 2 double on

*** News 10/15/97 -- Change to $myservers() (Colten)
	If you pass any argument to $myservers(), then a list of refnums to
	open servers will be returned.  The existing behavior of returning a
	list of NAMES to open servers is retained if you pass no arguments.

*** News 10/15/97 -- New built in function, $isconnected(refnum) (Colten)
	This function returns 1 if the specified server is currently connected
	(eg, you have been successfully registered).  If refnum is -1, then
	from_server is assumed.

*** News 10/08/97 -- New /window command, /WINDOW SCRATCH (ON|OFF)
	This allows you to designate a window as being a "scratch" window.
	A 'scratch' window is one that allows direct addressing, and it does
	not scroll.  The support *just barely works*.  That means if you 
	do anything i didnt expect, you probably will crash the client.
	Let me know if you find any ways to confuse this so i can fix them.

	To send text to a given line on a scratch window, use /xecho with
	the -line flag:
		/xecho -window <scratch refnum> -line <line number> text

	/XECHO -LINE will whine at you if:
	* If the window you specify is not a scratch window.
	* You dont specify -window before you specify -line.
	* If the line number you specify is out of range for the window.

	Suggestions/observations:
	* Scratch windows dont deal with being resized very gracefully.
	  I suggest you do /WINDOW NEW SCRATCH ON FIXED ON  so that it cant be
	  resized on you.  You should assume (at this time) that if the window
	  gets resized, the contents are *undefined*.  That will probably be
	  improved in the future.
	* Scratch windows probably should not be attached to servers.  Use
	  /WINDOW DISCON to make sure they are unbound.
	* There is NO SCROLLBACK on scratch windows.  When you overwrite a 
	  line, it is gone forever.
	* If you "unscratch" a window, then it is implicitly /CLEARed.
	  This to ensure that the status of the window after being returned
	  to normal will be coherent.  The previous contents of the window
	  will be retained in the scrollback buffer.
	* If you do a normal /echo to a scratch window, it will use the line
	  below the most previously output line.  When you reach the bottom
	  of the window, it will go back to the top.

*** News 10/08/97 -- New /window command, /WINDOW DISCON
	This allows you to specify that a window is not to be bound to any
	specified server.  That means the window will NEVER recieve output
	from any qualified server event, unless you specifically use /xecho
	to redirect output to that window.

*** News 10/01/97 -- IMPORTANT! SEMANTIC CHANGES! IMPORTANT!
	The $winchan() functions and $chanwin() functions have been reversed.
	That is to say, $winchan() now takes a WINDOW REFNUM argument, and
	$chanwin() takes a CHANNEL NAME argument.  This is the reverse of
	what it has been before, and the change was done in order to keep the
	semantics of built in functions coherent  (eg, $chan*() functions
	always take a channel name, and $win*() functions always take a window
	refnum.  A /set has been provided to let you have the old semantics.
	     *** BUT PLEASE MAKE A PLAN TO CONVERT YOUR CODE TO THE ***
	             *** NEW SEMANTICS. PLEASE PLEASE PLEASE. ***

	To let you know if the new semantics or old semantics are in effect,
	the new $info command, ``$info(w)'' command will return 1.  So add
	something like this in your script to get compatability:
			if (info(w) == 1) { set winchan_hack ON }

	IMPORTANT!!!
	The value of $info(w) will change to 2 (two) when this hack has been
	phased out in some future version. *** PROGRAM DEFENSIVELY. ***  Plan 
	to do something reasonable if $info(w) returns a value of 2. (This
	means that the above /set is not available, and the new semantics are
	irrevokably in place.)  I promise that the old semantics will not be
	irrevokably phased out before EPIC4pre5, and that the old semantics
	will most likely be phased out at the time of the production release
	(EPIC4.000)  Please plan ahead for this.

	If all of this ends up being unreasonably complicated or a burden,
	then let me know, and i'll work something out with you.

*** News 10/01/97 -- New /set, /SET WINCHAN_HACK
	This is a *temporary* feature to help ease the transition for those
	who have a substantial amount of code that depends on the semantics
	of $winchan() being "backwards" (eg, that it takes a channel argument
	rather than a window argument).  When this is turned on, $winchan()
	and $chanwin() will have their "traditional", or "reverse" semantics.
	When this is turned off, they will have their "modern", or "correct"
	semantics.  THIS SHOULD NEVER BE TURNED ON UNLESS YOU REALLY NEED IT.
	Please do NOT write new code with the old semantics, it will only
	encourage the broken semantics.

*** News 09/26/97 -- New /on, /ON REDIRECT
	This hooks any time something is about to be redirected to someone.
	If you hook this silently, the redirect will be suppressed.  This 
	would then be useful to filter out stuff that you might not want
	to have redirected.
		$0  - the target to whom the text is about to be sent
		$1- - the text itself.

*** News 09/20/97 -- New function $servernick()
	You give it a server refnum, it gives you your current nickname
	on that server.  This function has been around a while, it just
	never got documented.

*** News 09/20/97 -- #define CONTROL_C_COLOR now in $info(o)
	It is represented by the letter 'c'.

*** News 09/20/97 -- New built in function $chanwin()  <THIS HAS BEEN CHANGED>
	This is the reverse of $winchan().  You give it a window refnum or
	name, and it gives you the current channel (if any) for that window.

*** News 09/20/97 -- New built in function, $winnicklist() (Colten)
	You pass it either a window refnum or a window name, and it returns
	the nicks that are currently on the "nick list" for that window.
	(See the help for /window add for info about the nick list).  If
	you dont specify any arguments, the current window is presumed.

*** News 09/20/97 -- New /set, /SET NO_CONTROL_LOG (Peter Evans)
	When this is /set to on, the client will not output any nonprintable,
	nonspace characters to the log file.  When off, it will have the old
	behavior (dumping everything right to the log as it appears on the
	screen.)

*** News 09/20/97 -- New /on, /ON WINDOW_CREATE
	When a window is created with /on window, then this new hook will
	be activated.  You may assume that any "window" commands that you
	specify in the body will work for the window being created.  This
	is intended for setting "defaults" for windows, like so:

		/on window_create * window double on

*** News 09/18/97 -- Comprimise for /SET STATUS_MAIL
	Because there has been some controversy over the inability to "get
	rid of that damned colon" for %M in /set status_mail, i have agreed
	to take it out.  However, for those of us who like to have 
	/set mail 1, we wont be left with a "(Mail: )".  There is a new
	script called 'setmail", that traps /set mail, and offers two new
	/set's, /set mail_format1, and /set mail_format2, for controlling
	how the mail should look like when /set mail 1 and /set mail 2
	is engaged (respectively.)  I would just suggest that you load it
	in your .ircrc (before you /set mail) and see what i mean.

*** News 09/18/97 -- New error tracking feature
	If an error occurs while you are loading a script, then the client
	will tell you about where the error occured.  I say "about", because
	the client doesnt actually try to run /load'ed commands until it 
	reaches the next valid command, so the command that is in error is 
	always the first one BEFORE the line stated.  As well, when you
	attempt to /assign to an invalid variable, it will tell you what the
	invalid variable name was, so you can hunt it down.

*** News 09/16/97 -- New /set, /SET STATUS_NICK (|Rain|)
	This is a control for how your nickname will appear in the status
	bar.  Following convention, you may use %N in the /set to determine
	where the nick should be placed.  The default is just %N by itself,
	which yeilds the traditional behavior.

*** News 09/11/97 -- New function $nohighlight()
	This function will take any arbitrary input and will convert all of
	the highlight characters (^B, ^V, ^_, ^O, ^C, and ^[) to a "printable"
	format (in reverse).  This allows operators to find hidden attributes
	in channel names that would otherwise be invisibly stripped.

*** News 09/10/97 -- New function, $deuhc()  (Peter Evans)
	This function will remove "*!" or "*!*@" from the front of the
	specified pattern, if it is present.

*** News 09/09/97 -- New /window argument, /WINDOW SKIP (ON|OFF)
	You may now mark that a window should be 'skipped' when you are
	moving to the NEXT_WINDOW or PREVIOUS_WINDOW.  The only way to
	"land" on that window then is to use /WINDOW GOTO or such.  This
	was intended to be used for windows that are not to recieve input,
	such as your OpView window.

*** News 09/09/97 -- New command, /SHOOK <type> <args>
	This is a hokey name, i admit.  Anyone got a better name for it?

	This function allows you to arbitrarily synthesize /on events with
	your own arguments.  This is intended to retro-fit new problems onto
	old solutions.  For example, you could do:

		/on ^public "% &KILLS Received kill message *" {
			shook KILL $2-
		}

	or something like that, to make your normal /on kill handlers to be
	executed while youre on an ef2.9 server.  Or so is the idea.  Let
	me know what you think of this, and if its broken, say so.  
	**** BE WARNED ****
	Code in on hooks that depend on side effects (like $userhost()
	being set) will be *utterly confused* by these synthetic events, and
	will probably get wildly incorrect results.  You yourself are 
	responsible for making sure that this doesnt happen to you!

*** News 09/09/97 -- New /window argument, /WINDOW FIXED (ON|OFF)
	You may now "fix" a window as being a given size.  When a window is
	"fixed", it will never be shrunk or grown when you shrink or grow
	another window, and it is immune to /window balance.  You can still 
	directly shrink or grow a fixed window.  The only exception to this 
	is if all windows on a screen are fixed, and the screen is resized, 
	the bottom window of the screen will be forcibly resized, although it
	would still be "fixed". This was all done so that the OpView window 
	can be set to 4 lines and it will stay that way unless you explicitly 
	change it.

*** News 09/09/97 -- New /set, /SET BANNER_EXPAND
	If this is set to ON, then the contents of /SET BANNER will be
	expanded ($-substitution will occur) before it is used.  If it is
	OFF, then the value of /SET BANNER is used literally.

*** News 09/05/97 -- Changes to how windows are resized
	When you resize your display, then the proportion of window sizes
	between the non-fixed windows will be retained.  That is to say,
	if you had two windows, sizes X and 2X, and you change the screen
	so that the first window has size Y, then the second window will have
	about size 2Y.  When you rebalance your windows (with /window balance),
	all the nonfixed windows will be set to approx.  the same size.
	Fixed windows will not be touched. 

*** News 09/05/97 -- Now friendlier with job control
	If you want to suspend the client but are worried about it pinging
	out, you should be able 'bg' the client now, and it wont stop due
	to (tty output) or mangle your screen.  You will need to stop the
	client (with 'kill -STOP %1') before you 'fg' it again, or the client
	wont accept input.  This is not a bug, this is a job control problem.
	This should satisfy anyone who still pines for 'ircserv' ;-)

*** News 09/04/97 -- New /on, /ON DCC_LIST
	Many people have asked for an easy way to reformat the DCC output.
	So this new on will output the 8 most interesting capabilities.
	Im open to adding new ones, just let me know.  The 'banner' of the
	list is output with $0 being "Start" and the rest of the fields being
	filler.
		$0  - The type of DCC connection
		$1  - "1" if encryption is on, "0" if its not  (future exp)
		$2  - Nickname of the peer
		$3  - Status of connection
		$4  - $time() when connection established, 0 if not connected.
		$5  - Size of the file transfer, 0 if not applicable
		$6  - Number of bytes transfered, 0 if not connected
		$7  - Description field (usually the full filename)

*** News 09/04/97 -- Some new behavior for your display
	A lot of the code that handles your display has been rewritten.
	The most important changes are:  When you scroll backwards, you will
	get the lines that actually appeared on your screen, not the lines in
	your lastlog.  Those lines are output at the width they were split
	for, and if your screen is a different size, they wont be re-split.
	This is not a bug.  Do not report it as a bug.  When you are in
	scroll mode, the screen will not automatically scroll for you if
	the display reaches the end of the screen.  Instead, that stuff is
	put into the hold buffer.  You can use the SCROLL_FORWARD key to go
	forward a page *even into the hold buffer*, and all new output will
	appear as long as that output doesnt make the screen scroll.  Use the
	SCROLL_END key to go back to the "normal" scrolling behavior.
	You cannot turn HOLD_MODE off if youre in scrollback mode.

*** News 08/27/97 -- New /set, /SET CLOCK_FORMAT
	This /set, if set, controls how the client should represent the
	current time on your status bar, among other things.  Rather than
	have the representation be hardcoded in the client, this will give
	you the flexiblity to have the time displayed just however you want.
	By default, this /set comes UNSET.  When this is unset, the more
	traditional behavior of /set clock_24hour determines how the time
	should appear.  If you /set clock_format, it *overrides* the setting
	of /set clock_24hour.

*** News 08/25/97 -- New built in functions, $substr() and $rsubstr()
	These functions allow you to look for a string inside of another
	string.  It takes two or more arguments, with the first argument
	being the string to look for (use double quotes if it has spaces
	in it) and the rest of the arguments being the words to look in.
	(You dont need quotes for this arg.)  It returns -1 if the string
	could not be found, or returns the offset (counting by zero) where
	the string may be found in the source.  $rsubstr() looks backwards
	from the end of the string rather than from the front.

*** News 08/22/97 -- New flag to /XECHO, -B
	The /XECHO -B flag instructs the client to prepend the current value
	of BANNER_VAR before the line to be output.  This allows you to output
	something with a banner without having to worry about BANNER_VAR.  If
	you have have /set show_numerics ON and the hook happens to be a 
	numeric hook, then the numeric will be output.

*** News 08/20/97 -- Limits on /echo now removed
	There have always been limitations to the size of lines being handled
	for display.  Recent limits have been 2048 characters or 40 lines to 
	the screen, and 20480 characters to the lastlog or to a log file.  
	All of these limitations have now been eliminated, and any size line 
	(within the confines of available memory) can be handled without 
	excuses.

*** News 08/19/97 -- Handling of ^C colors, new set, /SET CONTROL_C_COLORS
	^C colors (eg, ``mIRC colors'') are now handled in the output stream.
	Whether or not anything is done with them is controlled by a compile
	time option (which is further controlled by a /set).  If you #define
	CONTROL_C_COLORS, in config.h and you /set CONTROL_C_COLORS ON, then 
	they will be parsed and converted into colors.  If you #undef it, 
	then the value of the /set will not matter, colors will never be 
	honored.  Sorry, but the "old" behavior of leaving them unchanged in 
	the output stream is no longer available.  You either must honor them 
	or you must have them filtered out.  (I doubt anyone will have a 
	problem with this.)

	By default, the #define is #undef'd, and the /set is turned off.
	The defaults will never change in future releases unless there
	is overwhelming public pressure to change them.

*** News 08/17/97 -- New operator, #~ (prepend operator)
	This is similar to the #= operator, which appends text to a variable.
	This operator PREPENDS text to a given variable.

*** News 08/17/97 -- New flag for /timer, -REPEAT
	The /timer -REPEAT flag takes one argument, which is given to be
	the number of times that the specified action should be repeated
	(every specified interval).  The /timer -LIST flag reflects the
	number of events left to be scheduled.

*** News 08/17/97 -- New script, "MOTD"
	This script re-implements client-side MOTD files.  It uses the two
	below functions.  Its only provided for those of you who miss this
	feature.

*** News 08/15/97 -- New function $ftime()
	This function returns to you the "mtime" of a given file (see stat(2))
	In ordinary words, this was the last time that the file was modified.
	It is the time that you see for an ordinary "ls -l" listing.  It is
	returned in seconds since the epoch, so youll need to $strftime() or
	$ctime() to convert it to human format.

*** News 08/15/97 -- New function $irclib()
	This function is only useful if you need to know where the IRCLIB
	is.  An example of its use is the "motd" script.

*** News 08/15/97 -- New function, $winbound()
	This function takes one argument.  If the argument specified is the
	name or refnum of a window, then the name of the channel bound to that
	window (if any) is the return value.  If the argument specified is
	the name of a channel, then the window refnum to which it is bound
	(if any) is the return value.  In case of ambiguity (if you have a
	window name that looks like a channel name), the window takes
	precedence over the channel.

*** News 08/08/97 -- Notify can now do userhost lookups automatically
	Most people (everybody?) do a /userhost request in their 
	/on notify_signons, and that has been distressing to many servers,
	because when you connect there is a flurry of ISON and USERHOST traffic
	and many people have gotten trapped in "flooder mode".  Some have 
	tried, with limited success and much agony, to batch up userhost
	requests to help.  In the interests of cutting down on client->server
	traffic needed for NOTIFY events, the client can now automatically
	fetch userhosts on your behalf.

	To turn on this feature, /set NOTIFY_USERHOST_AUTOMATIC on.  When
	this is ON, the client will dispatch userhost queries for everybody
	who is 'triggered' for that minute.  It does intelligent batching,
	so you wont be flooding the server as much.  This is ideal for when
	you initially connect and your flurry of ISONs and USERHOSTs can
	sometimes make the server hate you.  If this is OFF, then the 
	traditional behavior applies as normal.

	To accomodate this new feature, /on NOTIFY_SIGNON now takes two
	arguments:  The new second argument is the userhost of the person
	signing on.  If you have the automatic userhost feature turned off,
	the second argument will be empty.

*** News 08/08/97 -- New /set, /SET XTERM
	You may now use this set to specify what "xterm" program you want
	to use for /window create.  Note that this can be overriden by the
	XTERM environment variable.  This was done on purpose, so that those
	who have been setting the XTERM environment variable wont be suprised
	by the /set's "default".

*** News 08/08/97 -- New servcmd support (SQUERY, SERVLIST, WALLCHOP)
	The client now natively recognizes the SQUERY, SERVLIST commands
	from av2.9 servers, and the WALLCHOP command from u2.10 servers.
	I didnt implement WALLCHOP in the same way as contributed, so I
	might get yelled at and change the implementation soon. ;-)

*** News 08/08/97 -- Can send multiple files with /dcc send
	You may now specify more than one file with /dcc send nick <file>.
	I guess thats useful in conjunction with $glob(*) or something.
	Each request is sent out in a seperate PRIVMSG (obviously), so
	you should be careful not to send out TOO many files at a time...

*** News 08/08/97 -- New /on, /ON KILL
	This should simplify a lot of people's lives.  When you recieve
	one of those messages "*** Notice -- Recieved KILL message for ...",
	/ON KILL will be hooked with the following arguments:
		$0  - The server who sent the message (your server)
		$1  - The hapless victim
		$2  - The bastard who did the kill
		$3  - The server path to the bastard
		$4- - The reason for the kill

*** News 08/04/97 -- New /on, /ON OPER_NOTICE
	The server uses "sendto_ops" to send an notice to all operators.
	The client now automatically deals with this.
		$0  - the server that sent you the notice
		$1- - the text of the message AFTER the *** Notice --- part.


*** News 07/28/97 -- New built in function, $uhc()
	$uhc() takes a single argument, which is a partially constructed
	nick!user@host pattern, and returns a fully qualified nick!user@host
	pattern.  That is to say, any parts that are missing are filled in.
	This is a front end to cut_n_fix_glob(), which is the function that
	"fills in" your arguments to /ignore, so this works identically as
	the nick!user@host argument to /ignore.

	$uhc(nick)		returns		nick!*@*
	$uhc(user@host.com)	returns		*!user@host
	$uhc(*.host.com)	returns		*!*@*.host.com


*** News 07/28/97 -- New /SET, DO_NOTIFY_IMMEDIATELY
	If this is set to ON, then when you do the /notify command, an
	ISON will be immediately dispatched (the usual behavior.)  But if
	this is set to OFF, then the ISON is not immediately performed, and
	you will have to wait until the top of the next minute to see if 
	the person is on.  This is intended to be used in your .ircrc, 
	because some users have a zillion people on notify, and they end up
	sending enough lines to the server when they connect that they activate
	the server's flood control mechanism, then the server starts slowing
	you down, and thats just a big hassle.

*** News 07/28/97 -- New /SET, NOTIFY_INTERVAL
	This one is a bit tricky.  The value of this set is taken to be
	a suggestion about how frequently NOTIFY checks should be performed.
	This does not ever make the client check at any time other than the
	top of every minute, but it is possible to tweak this variable so
	that some minutes will be skipped.  For most sane people, setting
	this to be a multiple of 60 is advised.  Setting this to less than
	60 does not make it check more frequently than once a minute.

	/set notify_interval 60		yeilds the "customary" behavior
	/set notify_interval 120	makes it only check every 2 minutes
	/set notify_interval 150	makes it check only twice every 5 mins
					at minutes 3 (180s) and 5 (300s).


*** News 07/28/97 -- Two new /SETs, AUTO_REJOIN_DELAY and AUTO_RECONNET_DELAY
	These two control the amount of time that must pass between when
	you are kicked or killed, and the time that you attempt to rejoin
	or reconnect.   Reconnect is probably not totaly intuitive, as it
	attempts to bind the server to the current window *at the time that
	the timer expires*, not the window that was disconnected from.
	Arguably, this is a bug, and may be "fixed" in the future.  The 
	implementation for these is kind of hairy.


EPIC4pre1

*** News 06/18/97 -- New feature, "CPU SAVER"
	Some of you have asked me for a way to make ircII use less CPU
	especially when you are idling and detached.  Every minute ircII
	sends out your notify list, parses the replies, checks the clock,
	updates your status line, and other various housekeeping duties.

	But when youre detached, you dont really care if your notifies
	dont get processed promptly every 60 seconds, you dont care if
	the clock is accurate every moment, and you would prefer that
	the client not go to the trouble of wasting CPU time just to
	keep track.   This wont help if youre on busy channels, or
	logging wallops, only if the only traffic youre getting are
	the pings from the server.

	There are several new sets:
	/set CPU_SAVER_AFTER -	If you havent typed anything after
				this many minutes, CPU saver mode activates.
	/set CPU_SAVER_EVERY -	If cpu saver mode is on, the client will
				only do housekeeping every this many minutes.
	If either of these two are 0, then cpu saver mode cannot be
	activated, and the "traditional" behavior will apply.

	If youre interested in seeing when the client is in cpu saver
	mode, there is a new status line expando, "%L".  There is also
	a new /set, /SET STATUS_CPU_SAVER which determines how the %L
	expando will look when it is activated (just like STATUS_MAIL)

	There is also a new keybinding, "CPU_SAVER", which if bound to
	a key will activate cpu saver mode without having to wait.

	I did this on a whim, mostly because i havent done any features
	in a long time that have required this many changes, and I wanted
	to have some fun, so i just let loose with this and went berzerk
	If i forgot some obscure feature with this, let me know! ;-)


*** News 05/31/97 -- New environment variable honored, ``IRC_SERVERS_FILE''
	Someone had suggested something close to this -- they wanted a 
	way to specify a SERVERS_FILE as an absolute path rather than
	having it in their irclib directory.  So you can now set this
	environment variable to an absolute path where you would like the
	client to get your server list from.  This overrides any
	built in SERVERS_FILE setting, if appropriate.

*** News 05/31/97 -- Semantic clarification for /TOPIC
	Due to the confusion between the very common ``topic'' alias
	and the built in TOPIC command syntax, I have expanded the
	built in TOPIC command to do everything that everyone expects
	TOPIC to do, towit:

		/topic This is a topic		(set current channel topic)
		/topic * This is a topic	(same thing)
		/topic				(see topic on current chan)
		/topic *			(same thing.)
		/topic #chan This is atopic	(set topic on #chan)

	I strongly advise you to throw away your /topic alias unless it
	does something up and beyond this.  If it does, let me know, and
	I can include its semantics in the client, if its reasonable.

*** News 05/19/97 -- All new syntax for /IF
	As if ircII wasnt becoming PERLish enough, ive modified the
	/IF command to be more perl-like:  You may now use the 
	"elsif" modifier to indicate a blocked if command, eg:

		if (....) {
			....
		} elsif (...) {
			....
		} elsif (...) {
			....
		} else {
			....
		}

	(The final "else" is optional.)  Note that the placement of
	the "elsif" and "else" modifiers is critical:  They MUST be
	on the same line as the closing } to which they modify, or the
	/LOAD parser will get totaly confused and munge the whole
	thing up.  This is a limitation that applies only to how
	scripts are loaded -- sorry.

*** News 05/09/97 -- New function added, $findw(word ...text ...)
	This addresses the two most common complaints people have
	about $[r]match():

	1) This function returns the WORD NUMBER of the word
	   in text, such that $word($findw(word .text.)  .text.)
	   returns "word".  It returns -1 if word is not found.
	2) This function does not do anything screwy about quoting
	   or unquoting your input.  It just looks for the word,
	   thats it.  No pattern matching, no frills, and its fast.

*** News 05/09/97 -- Semantic weakness closed
	There was a semantic weakness with respect to the word count
	and strlen modifiers ($#var and $@var) such that if you did 
	not provide a variable to modify, an unuseful value was returned.
	Since this was likely to cause confusion, these cases are now
	defined to implicitly assume $* as the default argument. eg:
		$#   all by itself is now the same as $numwords($*)
		$@   all by itself is now the same as $strlen($*)
	These are *NOT* built in functions.  The # and @ characters
	still are considered to be modifiers.  The weakness was that
	these modifiers did not return meaningful values when a value to
	modify was not provided:  This change stipulates a useful default.

	In a seperate, but related topic, a change was also made to
	the math parser to honor this change in that context.  That
	is to say,   @ foo = #   now assigns to `foo' the number of
	arguments in the current context.  NOTE THAT THIS IS NOT A
	FUNDAMENTAL CHANGE!  We're just saying that the default value
	for the # and @ modifiers is $*.  This is not a "function"!

*** News 05/05/97 -- New flag to /WINDOW,  /WINDOW LASTLOG
	/WINDOW LASTLOG lets you individually tweak the size of a 
	window's lastlog.  This was introduced because this used to
	be the default beahavior of /set lastlog, but now that /set
	lastlog is a global action, we still need a way to do this...

*** News 04/29/97 -- New /set, /SET QUIT_MESSAGE
	This will be your default quit message.  If you do something
	stupid, like unset this completely, the client's version will
	be used as the default.  Of course, any argument you pass to
	/quit or /signoff will override this /set.

*** News 04/29/97 -- New /set, /SET DISPATCH_UNKNOWN_COMMAND
	If you type a command that is not recognized by the client,
	and youre one of those people who /never/ mistype a command,
	then you can /set this to ON, and the client will dispatch any
	otherwise unrecognized commands to the server.  Im sure your
	server admin will love you if you do this.

*** News 04/29/97 -- New built in function, $winchan() <THIS HAS BEEN CHANGED>
	You give it a channel name, and an optional server name
	or server refnum, and it tells you what window refnum that
	channel belongs to. (Written by IceKarma)

*** News 04/29/97 -- New flag to /lastlog, /lastlog -max
	You can use /LASTLOG -MAX to specify the maximum number
	of matches you want to display, regardless of any other
	flags.  The default is no limit.  This was written by Sheik.

*** News 04/29/97 -- New functionality to /stack,  /stack set
	You can now do /STACK (PUSH|POP|LIST) SET just as you can
	already with ALIAS/ASSIGN/ON.  This was written by Colten.

*** News 04/21/97 -- New set, /SET BANNER
	The /SET BANNER variable now controls what special banner will
	be prepended to informational messages from the client.  Up until
	now, the banner has been three stars ("***"), and many people dont
	like that or want to be able to change it.  Now you can.

*** News 04/21/97 -- New on, /ON SET
	/ON SET is a way to extend the capabilities of the /set command.
	The /ON SET hook is passed two or more arguments, with
		$0  -- is the name of the variable to be set
		$1- -- is the value the variable will be set to.

	The main idea is the ability to offer /set variables that are
	not built into the client, and execute ircII code to handle that
	situation.  The other idea is the ability to suppliment or override 
	a built-in set variable when it is changed.

	If you hook /ON SET with the "SILENT" modifier ('^'), then the
	default action (if any) will not be taken.  This may mean that
	if the variable is not a built in variable, no error will be
	output.  If the variable IS a built in variable, then the variable
	will NOT be set after the fact.

	There are two ways to get around this:  You can either hook the
	/ON SET with the "quiet" modifier ('-'), or you can use the /set
	command inside of the /on set body.  If you attempt to /set a 
	variable that is already being parsed by /on set, you will NOT
	be offered the /on again -- it will directly set the variable:

	Example:
		on ^set "exec_protection off" {
			echo *** You cannot set EXEC_PROTECTION off!
			set exec_protection on
		}

	Or:
		on ^set "auto_rejoin_delay *" {
			echo *** AUTO REJOIN delay set to [$1] seconds.
			@ myscript.autorejoin.delay = [$1]
		}

	Then you could do:
		/set auto_rejoin_delay 4


*** News 04/17/97 -- $X now reflects your ``real'' userhost
	When you connect to a server, the client asks the 
	server what your userhost is, and then that value is
	reflected in the $X variable.  Different server connections
	may have different $X values.

*** News 04/17/97 -- Changed semantics for USERHOST command
	The userhost command will act the same as always, except
	under the following  condition:

	If you do /userhost <nick(s)> -cmd, and *every* nick so 
	specified is on one of the channel(s) you are on (so that 
	the client already has the userhost cached), then the -cmd 
	code will be executed immediately, and the client will *NOT* 
	ask the server for the userhost information.  You are 100% 
	guaranteed still to get accurate information, you will just
	get it without waiting for the server query!  The only caveat
	is that the AWAY and OPER fields will be specified with the
	false value.  If you currently use the /userhost command to
	determine if a person is an operator, keep reading.

	A new flag has been added to the USERHOST command, -direct.
	The -direct flag forces the client to do a server query for
	the specified nicknames, even if all of the nicknames are 
	known to the client beforehand.  This can be used to get
	accurate AWAY and OPER information.  This flag only makes sense
	with the -cmd flag, since userhost caching does not happen
	unless the -cmd flag is specified.

*** News 04/10/97 -- Two new operators =~ and !~
	The two operators are intended to be similar to the perl operators:
	The lhs of the operator is expected to be a plaintext string, and
	The rhs of the operator is expected to be a wildcard expression.
	The operator returns a true value if the pattern matches the text,
	and it returns the false value if it does not.

	This is intended to be a low-cost alternative to the $match() and
	$rmatch() built-in functions when the only thing you need to do is
	see if a string is matched by a pattern.  If you need to select 
	from more than one pattern, use $match() or $rmatch().

*** News 04/07/97 -- Asynchronous code may use local variables
	You may reference local variables in asynchronous code
	provided that you follow the rules that have already been
	laid down.  See the regress/kill script for more info.

	1) You have to use /bless in the asynchronous code in order
	   to have access to the underlying local variables.
	2) You have to use /wait in the synchronous code in order to
	   make sure the local variables dont go away.
	3) Make sure you clean up your own messes.  Asynchronous code
	   lying around, resulting in /bless calls when there is no 
	   context for local variables is not a good thing.

*** News 03/31/97 -- New argument to /WAIT command, ``for''
	Yes, ``for'' without a hyphen.  This command guarantees that
	the code given as the argument is executed synchronously.
	That is to say, if the arguments, when executed, sent something
	to the server, the command will not return until that request
	has been completed (as if a wait had been done.)  But if the
	code does not send a query to the server, no wait is executed,
	and the command returns immediately after the code completes.

	You can also mix and match ``wait for'' with regular ``wait''
	calls and the client will be able to figure out what to do
	without waiting extra more than it needs to.

	And as always, using /wait or /wait for and /redirect at the 
	same time is a bad thing (tm), so dont do it.  The result is
	undefined behavior (eg, what happens isnt my fault.)

*** News 03/19/97 -- About the new who/ison (and soon userhost) queues.
	The client now keeps a FIFO (queue) of who/ison/userhost queries
	you make.  It is *critically* important that you do not confuse
	these queues (due to the nature of these server queries, there is
	no way to have any reasonable error recovery). 

	The following behaviors are forbidden (but the client wont stop
	you from doing them)

	* Using /ON ^RAW_IRC to wedge the 303, 351, or 315 numerics.
	  If you do this, the queues wont be properly flushed, anyone
	  waiting on a proper reply wont be handled, and any further
	  reqeusts will be thought to be previous queries, and all hell
	  will break loose.
	* Using /QUOTE to launch USERHOST, ISON, or WHO requests..
	  If you do this, the client will whine at you when it gets
	  unexpected replies and will eat the information.  If you launch
	  another query before the invalid one finishes, all hell will
	  break loose.

	*** THERE IS NO WAY TO RECOVER IF YOU DO THESE THINGS ***
	If youre stupid enough to do it, you get what you deserve.

*** News 03/18/97 -- Expanded syntax for $userhost()
	For your convenience, the $userhost() function now takes arguments:
	You may specify one or more nicknames as arguments to the $userhost()
	function, and the corresponding userhost(s) for the nick(s) specified
	will be returned.  The string <UNKNOWN>@<UNKNOWN> will be inserted
	for any nicknames whose userhost is not known.  The nickname(s)
	specified *must* be people who are on channels you are also on for
	the current server!  (See the second caveat)

	* Caveat -- Because it can take time for the WHO query to complete
	after you join a channel, there is no definite way to know if
	$userhost() for a person on your channel will succeed or not.
	You should probably be prepared to launch a /USERHOST query in
	case of failure.  This deficiency may be altered or modified in
	the future to block if an otherwise valid request is made while
	a WHO query is still pending.

	* Caveat -- This function will never be modified to launch a server 
	query.  That means that this function will only ever recognize those 
	nicknames that are common to channels you are on.  If you want to find 
	the userhost for other users, you should use the /userhost command.
	It is expected that the /userhost command will soon be able to
	take advantage of the userhost caching (but it does not yet.)

*** News 03/18/97 -- Expanded syntax for /who
	For your convenience, the /who command now takes two new arguments:

	-line {...}	The code inside the braces will be executed for each
			line returned by the who query.  The arguments are
			exactly the same as for the /on who hook, except that
			this flag is *guaranteed* to override the default
			/on who, and will also *go away* when the current who 
			query is finished.

	-end {...}	The code inside the braces will be executed at the
			end of the who query, when the 315 numeric is parsed.
			The arguments are:
				$0 - server queried
				$1 - body of the query
			This is guaranteed to to override the default /on 315,
			and will also go away once the current who query is
			completed.

	These command bodies will be executed asynchronously, which means
	that you must assume that the enclosing code scope will complete
	before the code is ever executed (e.g., the same rules as for the
	/userhost -cmd flag.) unless you use /wait.  You are encouraged
	to not use /wait if you can figure a way around it.  Code that is
	not concerned with being stricly backwards compatable with the
	stock client is ***strongly encouraged*** to use these new flags.

	Underlying this change is a new re-entrant WHO queue, which allows
	you to launch more than one WHO query simultaneously.  No longer do
	concurrent WHO requests have to wait for previous queries to complete,
	and no longer (with the above flags) does the caller need to put the
	WHO command in a /stack-/on-/wait wrapper.

*** News 03/17/97 -- /on 312, /on 319 changes
	All of the whois numerics: 311, 312, 313, 314, 317, 318, 319
	now pass $0 as the server name and $1 as the target of the 
	whois query.  All the rest of the arguments are passed as $2-.
	Previously, some of these numerics passed the target as $1,
	but not all of them.  I changed it this way for consistancy.

*** News 03/17/97 -- New: /ON STATUS_UPDATE and $status()
	/ON STATUS_UPDATE is hooked whenever any of the status lines
	for any of the visible windows changes **and you are in dumb mode**:
	   $0  is the refnum for the window whose status line has changed
	   $1  is the status line in that window that has changed
	   $2- is the actual status line for that window.

	$status() can be used to fetch the current status line for any
	visible window.  Invisible windows do not have their status lines
	updated, so they may be inaccurate.  This problem may change in
	the future.  Pass two arguments:
	   $0  is the refnum for a window
	   $1  is the status line

*** News 03/02/97 -- Local variables can now span entire array tree
	The syntax:
		``local x.y.''
	will define all variables in the subarray $x[y][...] to be
	implicitly considered local variables.  That is to say, the
	entire variable tree rooted at $x[y] is local to that scope.
	This was requested by a user for compatability with another
	client.  All local variables that are instantiated in this
	manner have the same scope as the explicit declaration above.

*** News 02/18/97 -- New built in variable, /SET CONNECT_TIMEOUT <seconds>
	Set this to the number of seconds you want your connect()ions to	
	block before they time out.  Note that this doesnt affect
	connect()ions that are already in progress, only those that are
	started after you set it.  The default is 30.

*** News 01/30/97 -- Expanded syntax for /LOCAL, /STUB
	You may now specify more than one name per command:  Each name
	will be treated independantly of all others.  The names must be
	seperated by a comma, and ***must not*** have any spaces between
	the commas and the names (that is, all of the names must form one
	logical word).  To declare both "foo" and "bar" as local vars,
			/local foo,bar

	You can do this to stub more than one alias/assign to the same
	file, as well:
			/stub foo,bar filename

*** News 01/29/97 -- New command /BLESS
	Currently this command ignores its arguments.  This may change
	in the future, so you should not get in the habit of supplying
	arguments to this function, for forwards compatability.

	The ''BLESS'' command is used to allow an asynchronous scope to
	access the local variables of the underlying synchronous context:

		alias foobar {
			local myvar $0
			userhost $1- -cmd {
				bless
				echo $myvar -- $0!$3@$4
			}
			wait
		}

	Note that the use of 'bless' must be paired with an appropriate
	call of ``wait'' in the synchronous context, or this wont work.

*** News 01/27/97 -- New flag for /EXEC, -direct
	Syntax:
		/EXEC -direct <commands>

	This simulates the effect that unsetting the SHELL variable
	causes, that is, the command is executed directly, without
	invoking a subshell.  This is appropriate for those who wish
	to execute a command passing untrusted data as an argument:
	With this flag, there would be no shell to interpret metacharacters,
	thus cutting down on that ability to have an accidental back door.

*** News 01/27/97 -- New command, /SETENV
	Syntax:
		/SETENV <var-name> <new-val>

	Sets the process environment variable <var-name> to the value of
	<new-val>.  This command is not neccesarily useful, nor is it 
	always appropriate, but it is provided for completeness.  There
	are several places in the code that can benefit from the ability
	to change environment variables after startup (eg, TZ).  It also
	allows you to set environment variables for /exec'd processes,
	and it is also possible to really foul things up if you mangle
	important environment variables such as DISPLAY and so on.

*** News 01/24/97 -- New, /ON ODD_SERVER_STUFF
	This is hooked whenever the server sends you something
	that the client doesnt recognize.  This is usually when
	you try to connect to a non ircd server, or you use some
	extension that isnt yet supported.  You are strongly 
	encouraged to use this interface to support new features
	rather than using /on raw_irc because this hook is only
	activated when something unexpected occurs (which is rare),
	rather than for every incoming line, as raw_irc does.

	$0  is the server that sent it to you, '*' if unknown
	$1- is the unrecognized command and its arguments.

*** News 01/22/97 -- /FE, /FOREACH use local variables now
	The "control variables" for the /FE and /FOREACH command
	now are local variables.  They will not disturb global variables,
	but they will destroy any local variables you have by the same
	name.  You dont have to explicitly declare the variables as local,
	it is done automatically for you.  Note that the normal warnings 
	for local variables still apply -- dont use any variable name that
	is the same as a global variable you might want to use later in
	the same alias/on. 

*** News 01/18/97 -- Support for local variables
	You may define a local variable with the following command:

		LOCAL <varname> [<value>]

	It mirrors the ASSIGN command.  Note that local variables act
	in exactly the same way as normal variables, except you cannot
	get rid of them.  You also do not (yet) have any way to get at
	a global variable that has the same name as a local variable, so
	choose wisely.  An example:

		alias foobar
		{
			local b			(declare LOCAL var $b as []) 
			assign a 4		(assigns 4 to GLOBAL $a)
			local a 5		(assigns 5 to LOCAL $a)
			eval echo $a		(outputs '5' -- local)
			@ a = []		(clears LOCAL variable)
			eval echo $a		(outputs '' -- local)
		}

	Restrictions:
	* You cannot have local aliases -- only local variables
	* You cannot save local variables.
	* You cannot (yet) access a global variable with the same name
	  as a local variable.

	This support is EXPERIMENTAL and still has some development to go 
	before it is mature.  


[This file truncated for EPIC4pre1... the rest of it is available via
 ftp at ftp.epicsol.org]

[[ Additional note ]]
People are apparantly looking in here to see where some of the features 
(like translation tables) went to.  Please refer to "doc/missing" before
reporting a removed feature as a bug.  Thanks.
