125) Why has OSF introduced Keysyms into Motif 1.1? They weren't

there in Motif 1.0.

Answer: From: ellis@osf.org

Virtual Keysyms are meant to provide a consistent keyboard model for Motif
applications running in a heterogeneous environment in which proprietary (i.e.
vendor specific) non-Motif applications may also be running.

First of all, for the sake of the rest of the readers, let's explain why this
is an issue:

It would be lovely if Motif's translation tables could just use the obvious
keysyms predefined by X.  For example, there are keysyms for XK_BackSpace,
XK_Delete, XK_Left, XK_Right, etc.  Shouldn't these be the ones that are used
in our translations?  Unfortunately, the problem is not so simple.  Some
specific examples:

   While most vendors bind XK_BackSpace to the key at the top right
   of the standard keyboard (often engraved with a leftwards
   pointing arrow), not all do.  In fact, some vendors (including DEC)
   bind that key to XK_Delete.

   While most vendors bind the arrow keys to XK_Up, etc, a number of
   vendors (including Sun, on some servers) bind them to function key

A simplistic solution would require the use of xmodmap to change the offending
bindings.  That would work swell in an all Motif environment.  However, OSF's
goal (not always perfectly achieved) is interoperability.  That is, we'd like
to make sure that both Motif and non-Motif programs can happily run in the
same environment.

It is expected that a vendor may have a wide variety of existing X-based
software that uses the keysyms as established by that vendor for specific
purposes.  It is expected that these applications may run at the same time as
Motif-based software.  Using xmodmap to change keysyms on the server side
could "break" the existing applications (or at the very least their
documentation) by making some keys unavailable, or by moving the location.

So, we chose not to use xmodmap.  By the way, though OpenLook uses a different
implementation (they recompile their virtual translation tables into actual
translation tables), they basically adopted the same approach, presumably for
similar reasons.

To work properly, the virtual keysym model we implemented depends on Xlib
finding XKeysymDB installed appropriately (which standard Motif installation
does).  This simply defines the keysyms (not the key they are bound to).  This
unfortunate piece of stupidity is necessary because MIT only includes standard
keysyms in keysymdef.h.  It should be said that our lives would be made easier
if MIT would also see fit to include registered keysyms in keysymdef.h as

Motif applications determine how to bind virtual to actual keys by looking for
either a resource or a property on the root window which describes what to do.
Note that this information is on the server side, so that all applications use
the same virtual bindings regardless of where they are running.  Mwm will
happily create the property if it finds a .motifbind file in your home
directory when it starts up.  (Actually, things generally work even if none of
this is done, since if all else fails, the Motif toolkit chooses a virtual
bindings table to use based on the identification of the server).

The actual implementation of virtual keys is made possible by a hook in the
Intrinsics.  Undoubtably, the implementation would be simpler and cleaner if
virtual key support was more directly supported by the Intrinsics.  We will be
exploring this possibility in the future.

  -- Ellis

Go Back Up

Go To Previous