Could have a size on objects. Then, a container would have a maximum size
that it could handle, and its size would be its base size plus the size of
everything in it. => Don't bother. It might add a bit of realism, but likely
doesn't do much for gameplay.

Could put a digital clock in the title bar of the MUD program. => Don't bother.
It doesn't add much and is kinda tacky.

Might be nice in the proving grounds to have:
    place to buy speed/strength/hits  (suggest prices each of 50000/50000/1000)
    very occasionally find gems lying around
    need place to sell them

If a picture is loaded that has no palette, it would often be useful to
    reset to the default palette at that time. Then again, if a scenario
    is consistently using a different palette, you wouldn't want to.
    So, let's not. The scenario can force a palette if needed.

Might want to add an 'ActionToString' builtin, to return a string containing
the text of a routine. The resulting string would be a pain to process,
however (essentially have to write part of a proc parser in MUD code), so
I don't think it would really be useful.

Could conceivably add a 'resynch' packet to the serial protocol. This could
be used to re-establish a connection that MUDAgent thinks is still there,
but that went away on the client end. This would only work on a direct line,
however, since MUDAgent normally pays attention to CD. It could also be a
big security hole - there is no way to check that the person reattaching
is the same one that detached.

In the build stuff, could have button-driven entries for destroy object,
move symbol, rename symbol and delete symbol. Decided not to - they are
less likely to be needed and can be potentially dangerous.

Some "spam" preventing stuff can be done. I think there is a clock tick
going off every second. So, when it does, check the number of inputs that
have come from a client in that second (keep a count, and reset on the
ticks). If it is over a threshold that is settable by a builtin call, then
mark that client as spamming. In that mode, send failing replies to any
request that needs a reply. For requests that don't (most), just chuck
them. Hmm. Perhaps just handle the ones that need replies - they are
wizard mode stuff mostly. Have another builtin-settable parameter which is
how long they stay in "spamming" mode. Need to set the first to a very high
value for the initial sourcing of the scenario (or anything similar).

Might also want to put limits on the number of rooms or objects that can
be created by a character using the build stuff. Give some kind of message
like "apply to SysAdmin for more". SysAdmin can just clear their counters.

Thought of having a 'FlashButton' graphics effect, so that when the user
uses a function key for moving, I could "flash" the on-screen image of
the corresponding button. I decided against this, since with a slow
connection or a slow server, the flashing would be too long after the
keypress to have the right feedback effect.

Had a note suggesting adding an AreaEnterCheck/AreaLeaveCheck action list.
I believe this was to be used when teleporing in/out of an area, to do
whatever cleanup was necessary. E.g. if a wizard teleports out of Questor's
office, no-one else can go in. Decided not to do this for now - the wizard's
will have to be careful.

Could add a flag to MUD and MUDAgent for "-reliable", which would indicate
that the serial connection is fully reliable. Then, don't need to use CRC,
ACK/NAK or sequence numbers. Decided not to do this. The gain would likely
not be that much, either in CPU usage or amount of data transmitted. Also,
any error or glitch encountered would hang the connection permanently.

Parser Extensions:
    Would like to be able to properly parse things like:

	take a picture of the book
	sit down on the bench		[currently kludged]

    This could be done either by allowing a multi-word separator
    ("a,picture;of") or by allowing multi-word verbs ("take,a,picture,of").
    Either requires some changes to the data structures used in memory and
    in-database for grammars and verbs.

    Put off until later, if ever. (How important is it?)

Out on the nets is the concept of "Tiny-Mail", where a player can mail
commands to a MUD and get the results back. Could do this for AmigaMUD.
Is there any point? (It would be VERY slow to play!)

Want some kind of more intelligent/realistic NPC. This could be a lot
of work. Have to think about it. Not for this release.

It might be possible to do some kind of mazewars type game with AmigaMUD.
Experiments show that this would not be feasible at 2400 baud, even with
effects routines. To do this, might want to increase the resolution of
'After' to 1/100 second instead of 1 second. Maybe later.

Could add a 'CreateThingWithSpace(thing parent; int propCount)', which
would create a thing in the database with space pre-allocated for 'propCount'
properties. Might reduce fragmentation. Probably not worth it.

The MUD client can end up temporarily suspended after typing a RETURN, so
that further typing works, but hitting RETURN again doesn't work. This is
carefully done in the text I/O code, but I don't quite remember the reason
for it. An extra level of recursion would happen if it were allowed, and
too many of these could cause problems. The situation is that the client
is busy doing something (like editing, or processing effects stuff), not
that the client is waiting for the server (input lines are not acknowledged).

Could have ANSI terminal output stuff. Then, players without the remote
client could get a bit more than just straight text. Would need routines
like ANSIOn, ANSIMove, ANSIClearLine, etc. One problem is the nature of
MUDs in general - output can come at any time. A possible approach to
handling this is to require that the ANSI emulator support a scrolling
region, which would be used for all output not specifically directed to
other parts of the screen. The scenario would have to make sure that any
non-scrolling output sequences leave the cursor at the bottom left of the
scrolling region. Descision: no ANSI support - we are trying to move to
graphics support - ANSI support doesn't help that goal and might even
hinder it.

Some MUDs have a "snoop" facility, which can let wizards or sysadmins
directly watch what players are doing. This includes displaying all of
their input and output. I could add such a facility to the AmigaMUD server
without too much trouble, but have decided not to. The only legitimate
uses of snoop that I can think of are a sysadmin watching for abusers;
and for wizards to watch people playing their creations, looking for
problems with them. Both of these can be handled by proper reporting.
The possible abuses of a snoop facility are many.

Under V1.3 it is possible to select multiple menu items at once, by holding
down the shift key. In doing so, it is possible to select any number of
the items in a set which are supposed to be mutually exclusive. Because
of the way Intuition reports these events, some will be missed if a given
item is selected more than once. If a program just follows the codes in
the menu events given to it, it can then be out of step with the checkmarks
in a set of items. This can be fixed by checking the state of the checkmark
flag bit before doing anything. I started to do this, but it was adding
too much code to handle a rare and not fatal case. Also, later versions
of the system handle this better. So, I do not handle this situation, and
do not plan to do so.

It might be possible to set things up so that reliable backups of the
database files can be made without shutting down the system. All structures
would have to be fully flushed, just as they would on a shutdown. It isn't
clear that this is done right now. Then, have a builtin that will lock
the database - it will not do any write's while locked - it will wait
for the lock to be released. Hmm. How is the lock released if the server
is waiting? Anyway, I've decided not to persue this - the sysadmin will
have to shut the server down in order to make a reliable backup. Another
possibility that I haven't fully tested is having the backups done from
the MUDServ program itself, using scenario code like:

    proc backup()void:
	Flush();
	Execute("copy MUD.#? backup");
    corp;
    backup().

That way, the actions are done atomically. Seems OK, but not fully tested.

Fighting in the dark is currently not allowed. Changing this, and keeping
darkness consistent (i.e. you can't see where you are or who/what is there
with you) would require a bunch of new checks and messages in the fighting
code, including changes to the 'hit' command (how does the player decide
who/what to hit if they aren't visible?). Could have 'hit it' hit some
random who/what in the same location. Anyway, not done for now.

There is currently no easy way to talk about parts of objects. E.g. if we
had a small box with a bunch of buttons on it, how do we set it up so
that the player can 'push red button'? One way is to have pickup/drop
routines on the box which add/remove invisible button objects to/from
the player's carrying list. Hmm. Just checking util.m - a much better
way is to add a 'push' p_oActWord to the box - that could then know about
the buttons and do the right thing. A limitation here is that a given
object only gets to have one 'actword'. A more general way that is not
implemented in this scenario is to use grammar properties on rooms and
objects, which must be checked before the main one. Note that this would
require some kind of pre-parsing of the input command to pick off a direct
object to look the command up on. This doesn't work well with the way
that 'Parse' handles multiple sentences in an input string. Perhaps what
is needed is a callout from 'Parse' that can check each individual
sentence in this way. Something to think about, but nothing done for now.

It would be possible to run the MUD client with no text window, if this
is requested by the scenario. In this mode, no text output should be
sent from the server to the client. This could be useful for games of
a slightly different flavour than the normal text MUD. All input would
be keypad or mouse, and all output would be graphics/sound. One COULD
fake a small text output window using the existing graphics commands and
some code in a scenario. There is currently no way to do text input in
this fashion - would require a 'raw' mode for other keyboard input, and
a handler for such raw keys. Decided not to persue this for now.

If there is no handler for mouse regions and/or buttons, it might be
nice to 'beep' if the user clicks in the graphics region. I think this
is a pretty minor detail, however, so I haven't done anything about it.

I'm somewhat bothered by the inconsistent nature of the current standard
scenario - it has areas of present-day modern-world stuff just down the
street from fantasy stuff like the Squirrel area, and D & D medieval
stuff like the Proving Grounds. There are lots of ways to blend these
together better. One would be to have a 'dimensional gate' in a building
on the street somewhere, and this is the opening to these other areas.
The 'teleporter gizmo' idea is another one. For now, I'm not going to
do either one. Perhaps if the mall/streets area gets expanded more, so
that there are more quests, and more stuff to explore.

More traditional D & D stuff could be added to the Proving Grounds area.
Player's could find gems/jewelry/artifacts pseudo-randomly (or in some of
the "special" areas), which they can redeem at a jeweler's shop or some
such. It would also be nice to be able to sell stuff back in the stores,
say for half-price or something. This could reduce clutter faster than
the monster-cleanup stuff. Nothing done for now.

Could write a builtin
    RunCheckers(thing room; property action checkProp; thing actor)status
that checks the room and all clients in the room for the prop. Any that
exist are called, until one returns other than 'continue'.
Not needed currently, so not done.

The text window on the MUD client should make use of 2.0 features and
routines when available. It should also open a text window consistent in
size and mode with the user's default settings (e.g. interlace). I'm not
doing this for now - I'm more concerned with functionality, and that this
version at least be useable on nearly all current Amigas.

In the current scenario, there are movement checkers in the current room,
on the direction in the current room, on the player, and in the destination
room. In some situations, it might be good to have some on other players
or machines in either of the rooms. So that, e.g. a monster could prevent
you from leaving a room. I haven't done this, partly because it would be
expensive in CPU time to do, and partly because most of the functionality
can be had by dynamically adding checkers as needed. E.g. the blocking
done currently by monsters in the proving grounds is handled by a single
checker added to the player as part of combat initialization.

I had written down that there should be a way to create machines with
empty or special names, such that they won't show up when you look around
the room. Not needed - just make them hidden and have null names.

It is desireable that a wizard be able to prevent others from adding to
their areas. This is done by making the rooms ts_readonly. This does not
prevent other wizards from making connections TO rooms, however. This
is prevented by the build code, but that isn't enough for wizards. It
would be possible to make the basic AmigaMUD interpreter not add a thing
as a property if that thing is ts_readonly and not owned by the effective
player. This could have bad consequences for purposes other than just
connecting up rooms, however. It would appear that either another basic
attribute of things is needed, or that the ts_ status values be expanded
to handle this. This needs more thought. For now, I will do nothing.

Need some way to delete stuff in bulk. E.g. if a SysAdmin wants to delete
a whole area, added either online with build commands/buttons or added
by sourcing files. How is the stuff to be identified? One way would be
based on the owner of the rooms. I think this could be done entirely in
the scenario, but I'm not sure. The code would have to build up a list
of the tables that the owning player has, and so look up in all of them
every thing, action, etc. that needs deleting. How does it delete
properties? There is no way to get the name of a property. This needs more
thought, and perhaps user feedback. Do nothing for now.

How useful is the concept of apprentices? Are they noticeably less able
to mess things up than full wizards? The only thing I can see currently
is that they can't build out from rooms marked ts_wizard, whereas full
wizards can. Thus, they can't mess up areas that are specifically so
protected. Is this worth it?

One way of identifying areas is to have an "id" on stuff. This could be
something that the player can set on themselves, and which is copied into
everything they create. Could do this with the build code - is that enough,
or is it wanted for wizard/apprentice direct building too? The ultimate
would be to make such an identifier be something in every symbol table
entry, and then have a builtin that deletes all symbols with a given id.
This could lead to recursion, so would require careful implementation if
it is to not loop, but still delete everything. Nothing done for now.

Something on an early list of things to do was to add some kind of
server supported trigger which would idle machines when there has been
no player on the MUD for a while. This would need a pair of different
idle/active actions on machines. I've decided not to bother doing this.
The current machines accomplish much the same thing by calling 'ClientsActive'
to see if there are any active players, and making their step timing much
less frequent if there aren't.

Another early desire was for a 'pageProc' on machines, and a 'page'
builtin to page players/machines. This isn't needed anymore. A player
can be paged using SPrint, and a machine can be triggered using 'ForceAction'.

Some MUDs have a 'gag' facility, which will prevent output from a given
client X from being sent to client Y. This is set by client Y, who can
do it for any number of clients. This is used to make conversations more
understandable when there are lots of people in a room, or to shut out
the cries of the mortals when a wizard is busy doing something. The proper
place to implement this is in the server, so that the unwanted output is
not even sent to the remote client. I have not done this, and I'm not sure
that it will be needed on AmigaMUD - most games will not have that large
a number of simultaneous players, and the wizards can always go somewhere
else to work, even if they have to create and block off that somewhere.
As a partial measure, the builder's guild has four separate meeting rooms
that are only enterable by builders.

The AmigaMUD programming language could be made a bit more complete by
allowing direct procs, as in:

    Me()@pProc := proc Print("Hello there world!\n"); fail corp.
    call(Me()@pProc, status)().

to have parameters and to have result types other than 'status'. This
would be conceptually nice, but I can't see any immediate need, and I
would rather not mess with that bit of code right now! The current
limitations can be worked around, and are certainly adequate for the
procs dynamically built via 'StringToAction' by the build code.

Could rig the shifted functions keys in the MUD client to be macro keys
that the user can program (e.g. via the MUD icon). Hey users! Is this
something I should bother doing? How should it be set up?

It should be possible to set up profiling of MUD code. Set up a timer
interrupt to go off often enough, and when it does, run through the
current interpreter traceback, incrementing a counter on each function
encountered. Alternatively, just increment the counter for the innermost
function on the traceback. Then have a command to dump the counts on
all procs in a given table. Is this of any real use? Not done for now.

If a change in prompt comes along when the MUD client input buffer is
not empty (i.e. in the middle of typing some input), things get messed
up a bit. I'm not too worried about this, so won't chase it.

If a remote server machine crashes, or the server is aborted, a MUD client
has no way of knowing, and will sit around waiting for a reply to whatever
requests it has sent. It might be useful to have a timeout applied to all
requests that need a reply (not readily known outside of mud.library!),
and ask the user if they want to abort if the timeout goes off without a
reply. I've decided not to bother with this. In most cases, the serial
protocol will time out and give up on one of its packets. The only case
where this won't happen is if only the MUDServ process on the server
machine bungs up, so that the MUDAgent continues to acknowledge packets
at the serial level. This would also happen with a text-only connection.
On a local connection, it should be fairly obvious if the server goes away.

It would be possible to handle a ^C typed on a client that is messing
around with wizard-level stuff. Have the server receive messages via an
interrupt, which just puts them on a queue of incoming messages. If the
handler sees a ^C message (special type), and it is for the client that
is currently executing in wizard mode, set the interpreter's shutdown flag
right away. Decided not to do this. The execution timeout limit can handle
most of the cases. A running system SHOULD have a reasonable timeout,
like 5 seconds or so. Also, in a standalone environment while developing,
execution of code under a long timeout can be stopped with a BREAK command
from a CLI.

It would be possible to trivially encrypt the database files. This would
make it harder for people to 'cheat' when trying to solve quests, etc.
Decided not to do this. Someone would soon produce a decrypting program
anyway.

It should be possible to allow the use of 'it' as a direct object. Just
retrieve the value of It() (complain if none) and grab its p_oName. Hmm,
that can't be done in the server, since it doesn't know about p_oName. So,
it would have to be done in all verbs in the scenario. This wouldn't be
too hard, but I'm not convinced the benefits are all that great - there
are likely lots of cases where it wouldn't do what the player expected,
hence most players wouldn't use it.

It would be nice if a future version of the system with higher resolution
(likely only usable on AGA systems) could have a 'MUD' that is compatible
with older servers. It would double all pixels sent by a lo-res scenario.
In order for this to work, it would have to send its resolution to the
scenario as the older, small size. To know it should do this, it needs
to know the server version before the EffectsInfo data is sent to the
server. Unfortunately, the EffectsInfo is currently sent in the first
message from a remote client to the server, so there is no chance to do
this. I guess it'll have to be a manual flag on 'MUD'.