DOC HOME SITE MAP MAN PAGES GNU INFO SEARCH PRINT BOOK
 

kbd(7)


kbd -- generalized string translation module

Description

The STREAMS module kbd is a programmable string translation module. It performs two types of operations on an input stream: the first type is simple byte-swapping via a lookup table, the second is string translation. It is useful for code set conversion and compose-key or dead-key character production on terminals and production of overstriking sequences on printers. It also can be used for minor types of key-rebinding, expansion of abbreviations, and keyboard re-arrangement (an example of the latter would be swapping the positions of The Y and Z keys, required for German keyboards, or providing Dvorak keyboard emulation for QWERTY keyboards). The kbdcomp(1M) manual page discusses table construction, the input language, and contains sample uses. It is intended mainly to aid administrators in configuring the module on a particular system; the user interface to the module is only through the commands kbdload(1M) and kbdset(1).

The kbd module works by changing an input stream according to instructions embodied in tables. It has no built in default tables. Some tables may be loaded when the system is first brought up by pushing the module and loading standard or often-used tables [see kbdload(1M)] that are retained in main-memory across invocations and made available to all users. These are called public tables. Users may also load private tables at any time; these tables do not remain resident.

With the kbdset command, users may query the module for a list of available and attached tables, attach various tables, and set the optional per-user hotkey, hot-key mode, and verbose string for their particular invocation.

When a user attaches more than one table, the user's hot-key may be used to cycle to the next table in the list. If only one table is specified, the hot-key may be used to toggle translation on and off. When multiple tables are in use, the hot-key may be used to cycle through the list of tables. [See kbdset(1) for a description of the available modes.]

In its initial state, kbd scans input for occurrences of bytes beginning a translation sequence. When receiving such a byte, kbd attempts to match subsequent bytes of the input to programmed sequences. Input is buffered beginning with the byte that caused the state change and is released if a match is not found. When a match fails, the first byte of the invalid sequence is sent upstream, the buffered input is shifted, and the scan begins again with the resulting input sequence. If the current table contains an error entry, its value (one or more bytes) is substituted for the offending input byte. When a sequence is found to be valid, the entire sequence is replaced with the result string specified for it.

The kbd module may be used in either the read or write directions, or both simultaneously. Maps and hot-keys may be specified independently for input and output.

The kbd also supports the use of external kernel-resident functions as if they were tables; once declared and attached (via kbdload and kbdset respectively) they may be used as simple tables or members of composites. To accomplish this, kbd understands the registration functions of the alp module and can access any function registered with that module. Further information on external functions and their definition is contained in alp(7). External functions are especially useful in supporting multibyte code set conversions that would be difficult or impossible with normal kbd tables.

Limitations

It is not an error to attach multiple tables without defining a hot-key, but the tables will not all be accessible. It is recommended that the user's hot-key be set before loading and attaching tables to avoid unpleasant side effects when an unfamiliar arrangement is first loaded.

Each user has a limitation on the amount of memory that may be used for private and attached tables. This ``quota'' is controlled by the kbd_umem variable described below. When a user that is not a privileged user attempts to load a table or create a composite table, the quota is checked, and the load will fail if it would cause the quota to be exceeded. When a composite table is attached, the space for attachment (which requires more space than the composite table itself) is charged against this quota (attachment of simple tables is not charged against the quota). The quota is enforced only when loading new tables. Detaching temporarily from unneeded composite tables may reduce the current allocation enough to load a table that would otherwise fail because of quota enforcement. To minimize chances of failure while loading tables, it is advisable to load all required tables and make all required composite tables before attaching any of them.

Configuration parameters

The master (or space.c) file contains configurable parameters.

NKBDU is the maximum number of tables that may be attached by a single user. The number should be large enough to cover uncommon cases, and must be at least 2. Default is 6.

ZUMEM, from which the variable kbd_umem is assigned, is the maximum number of bytes that a user (other than a privileged user) may have allocated to private tables (that is, the quota). Default is 4096.

KBDTIME is the default timer value for timeout mode. It is the number of clock ticks allowed before timing out. The value of one clock tick depends on the hardware, but is usually 1/100 or 1/60 of a second. A timeout value of 20 is 1/5 second at 100Hz; with a 60Hz clock, a value of 12 produces a 1/5 second timeout. Values from 5 to 400 inclusive are allowed by the module; if the value set for KBDTIME is outside this range, the module forces it to the nearest limit. (This value is only a default; users may change their particular stream to use a different value depending on their own preferences, terminal baud-rate, and typing speed.)

Files

/usr/lib/kbd - directory containing system standard table files.
/usr/lib/kbd/*.map - source for some system table files.

References

alp(7), kbdcomp(1M), kbdload(1M), kbdset(1)

Notices

NULL characters may not be used in result or input strings, because they are used as string delimiters.


© 2004 The SCO Group, Inc. All rights reserved.
UnixWare 7 Release 7.1.4 - 25 April 2004