[ < ] | [ > ] | [ << ] | [ Up ] | [ >> ] | [Top] | [Contents] | [Index] | [ ? ] |
E_INVARG
is raised.
E_PERM
is raised. If conn is not a currently-active connection,
then this function does nothing. Output is normally written to connections
only between tasks, not during execution.
The server will not queue an arbitrary amount of output for a connection; the
MAX_QUEUED_OUTPUT
compilation option (see section Server Compilation Options) controls the
limit. When an attempt is made to enqueue output that would take the server
over its limit, it first tries to write as much output as possible to the
connection without having to wait for the other end. If that doesn't result in
the new output being able to fit in the queue, the server starts throwing away
the oldest lines in the queue until the new ouput will fit. The server
remembers how many lines of output it has `flushed' in this way and, when next
it can succeed in writing anything to the connection, it first writes a line
like >> Network buffer overflow: X lines of output to you have been
lost <<
where X is the number of flushed lines.
If no-flush is provided and true, then notify()
never flushes any
output from the queue; instead it immediately returns false. Notify()
otherwise always returns true.
read()
simply returns 0 immediately.
If conn is provided, then the programmer must either be a wizard or the
owner of conn; if conn is not provided, then read()
may only be called by a wizard and only in the task that was last spawned by a
command from the connection in question. Otherwise, E_PERM
is raised.
If conn is not currently connected and has no pending lines of input, or
if the connection is closed while a task is waiting for input but before any
lines of input are received, then read()
raises E_INVARG
.
The restriction on the use of read()
without any arguments preserves the
following simple invariant: if input is being read from a player, it is for the
task started by the last command that player typed. This invariant adds
responsibility to the programmer, however. If your program calls another verb
before doing a read()
, then either that verb must not suspend or else
you must arrange that no commands will be read from the connection in the
meantime. The most straightforward way to do this is to call
set_connection_option(player, "hold-input", 1) |
read()
and other code that might suspend, and finally call
set_connection_option(player, "hold-input", 0) |
E_INVARG
if conn does not specify a current connection and
E_PERM
if the programmer is neither conn nor a wizard.
E_INVARG
is raised. If either string is currently
undefined, the value ""
is used instead. See the discussion of the
PREFIX
and SUFFIX
commands in the next chapter for more
information about the output prefix and suffix.
notify()
, connected_players()
, and the like) immediately behave
as if the connection no longer exists. If the programmer is not either a
wizard or the same as player, then E_PERM
is raised. If there
is no currently-active connection to player, then this function does
nothing.
If there was a currently-active connection, then the following verb call is made when the connection is actually closed:
$user_disconnected(player) |
It is not an error if this verb does not exist; the call is simply skipped.
E_PERM
is raised. If player is not currently connected, then
E_INVARG
is raised.
For the TCP/IP networking configurations, for in-bound connections, the string has the form
"port lport from host, port port" |
For outbound TCP/IP connections, the string has the form
"port lport to host, port port" |
For the System V `local' networking configuration, the string is the UNIX login name of the connecting user or, if no such name can be found, something of the form
"User #number" |
For the other networking configurations, the string is the same for all connections and, thus, useless.
E_INVARG
if conn does not specify a current connection and
E_PERM
if the programmer is neither conn nor a wizard. The
following values for option are currently supported:
"hold-input"
read()
or this connection option is set false,
in which case command processing resumes.
Processing of out-of-band input lines is unaffected by this option.
"client-echo"
"binary"
"flush-command"
$server_options.default_flush_command
; see the chapter on server
assumptions about the database for details.
{name, value}
pairs describing the
current settings of all of the allowed options for the connection conn.
Raises E_INVARG
if conn does not specify a current connection and
E_PERM
if the programmer is neither conn nor a wizard.
E_INVARG
if conn does not specify a current
connection and E_PERM
if the programmer is neither conn nor a
wizard.
read()
, notify()
, and boot_player()
. This object number
is the value returned by this function.
If the programmer is not a wizard or if the OUTBOUND_NETWORK
compilation
option was not used in building the server, then E_PERM
is raised. If
the network connection cannot be made for some reason, then other errors will
be returned, depending upon the particular network implementation in use.
For the TCP/IP network implementations (the only ones as of this writing that
support outbound connections), there must be two arguments, a string naming a
host (possibly using the numeric Internet syntax) and an integer specifying a TCP
port. If a connection cannot be made because the host does not exist, the port
does not exist, the host is not reachable or refused the connection,
E_INVARG
is raised. If the connection cannot be made for other
reasons, including resource limitations, then E_QUOTA
is raised.
The outbound connection process involves certain steps that can take quite a long time, during which the server is not doing anything else, including responding to user commands and executing MOO tasks. See the chapter on server assumptions about the database for details about how the server limits the amount of time it will wait for these steps to successfully complete.
It is worth mentioning one tricky point concerning the use of this function.
Since the server treats the new connection pretty much like any normal player
connection, it will naturally try to parse any input from that connection as
commands in the usual way. To prevent this treatment, you should use
set_connection_option()
to set the "hold-input"
option true on
the connection.
do_login_command
, do_command
, do_out_of_band_command
,
user_connected
, user_created
, user_reconnected
,
user_disconnected
, and user_client_disconnected
will be called at
appropriate points, just as these verbs are called on #0
for normal
connections. (See the chapter on server assumptions about the database for the
complete story on when these functions are called.) Point is a
network-configuration-specific parameter describing the listening point. If
print-messages is provided and true, then the various
database-configurable messages (also detailed in the chapter on server
assumptions) will be printed on connections received at the new listening
point. Listen()
returns canon, a `canonicalized' version of
point, with any configuration-specific defaulting or aliasing accounted
for.
This raises E_PERM
if the programmer is not a wizard, E_INVARG
if
object is invalid or there is already a listening point described by
point, and E_QUOTA
if some network-configuration-specific error
occurred.
For the TCP/IP configurations, point is a TCP port number on which to listen and canon is equal to point unless point is zero, in which case canon is a port number assigned by the operating system.
For the local multi-user configurations, point is the UNIX file name to be used as the connection point and canon is always equal to point.
In the single-user configuration, the can be only one listening point at a time; point can be any value at all and canon is always zero.
listeners()
. Raises E_PERM
if the programmer is not a wizard and
E_INVARG
if there does not exist a listener with that description.
unlisten()
). Each element of the list
has the following form:
{object, canon, print-messages} |
where object is the first argument given in the call to listen()
to create this listening point, print-messages is true if the third
argument in that call was provided and true, and canon was the value
returned by that call. (For the initial listening point, object is
#0
, canon is determined by the command-line arguments or a
network-configuration-specific default, and print-messages is true.)
Please note that there is nothing special about the initial listening point
created by the server when it starts; you can use unlisten()
on it just
as if it had been created by listen()
. This can be useful; for example,
under one of the TCP/IP configurations, you might start up your server on some
obscure port, say 12345, connect to it by yourself for a while, and then open
it up to normal users by evaluating the statments
unlisten(12345); listen(#0, 7777, 1) |
[ < ] | [ > ] | [ << ] | [ Up ] | [ >> ] | [Top] | [Contents] | [Index] | [ ? ] |