Re: Proposal for reporting remote control button presses to userland

[List Mail User <track%Plectere.com@localhost>]

>...
>Design question: should the receiver do the filtering, or should
>wsremoted?  
>
>               --Steve Bellovin, http://www.cs.columbia.edu/~smb
>
        It is quite depressing to see a discussion like this, with not even
a single party referring to the existing literature or any implementations
besides Linux (which certainly would not be a good reference here).

        While most implementations done by the workstation vendors (i.e. the
companies that did graphical interfaces, before MS Windows) and exist only
in proprietary code, there are plenty of man pages, client code (i.e. using
those interfaces) and even some papers published out there.  Twenty years ago,
"we all" read the Xerox PARC stuff.  Much is hiding in source code which was
donated by vendors to projects at various univerities (CMU in particular, as
well as MIT).

        Just seeing the lack of common jargon - e.g. confusing "buttons" with
"keys", is enough to know no one here has bothered to look.  Also, historically,
most vendors except SunOS/Solaris gave up on userspace daemons - just too many
context switches for even moderate event rates (e.g. accelerometer, pressure
sensor, strain gauge, etc.) - think secure shared memory transport (e.g. ring
buffers with validity checking by one party).  Good examples are Ultrix (very
limited event types, but very efficient), AOS (IBM's BSD4.3 source distribution
for the RT) and IRIX (though I wouldn't recommend using SysV Streams for the
input, it does make translators easy to write).
        Paul Shupak