🚀 go-pugleaf

In article <5504@bacon.IMSI.COM> jordan@IMSI.COM (Jordan Hayes) writes:
>Anybody have a good example of a real-life application that can be
>tolerant of dropped messages?

I'm not sure if it counts as "real-life", but as part of an undergrad
research project last year I modified the interface to an input device
so the machine hooked up to it could pass along the input data to an
application on a different machine. The speed advantage of UDP was a
big plus since the remote machine was doing some time-critical 3D
graphics stuff (ie: VR applications). The downside of possible lost
packets was not a big deal, since the input device used forces and
torques rather than absolute position. It was a case where under heavy
network load, it was better to toss the packets away than to buffer
them.  It also helped that an occasional dropped packet didn't make
much of a difference in the human time frame.

Of course, this was an unusual case. I agree that under conditions
where it's important to guarrantee delivery, you'd have to have some
pretty bizarre circumstances to warrant rewriting your own TCP on top
of UDP.

Just my two cents...
-- 
Dave                         <dnoble@rdrc.rpi.edu>
            Computer & Systems Engineering, class of '92 and a half
 Visual Technologies Program, Rensselaer Design Research Center, RPI, Troy NY
///// "Hardware, n.: The parts of a computer system that can be kicked." \\\\\