[K12OSN] 1 thin client = 4 terminals?

["Terrell Prudé, Jr."]

I wouldn't do it.  Here's why.  If you want to hook up more than one 
keyboard and mouse to a LTSP client, then you need at least two things:

1.)  some way to physically hook up the other monitors, keyboards, and 
mice, and
2.)  some way to separate each of the four sessions.

I just don't see how to do that with a LTSP client.  To illustrate this, 
let's first take the case of a regular "fat client" PC.  In a "fat 
client", you could install a bunch of serial ports in the thing, have, 
say, four VT100 terminals, and do it like that.  The VT100 terminal, in 
this case, is itself handling the display, as all terminals, including 
the ancient IBM 3278 mainframe terminals, do.  You *have* to do it this 
way otherwise it'd be like PCAnywhere in which someone logging into 
PCAnywhere will disrupt the operations of the person already sitting at 
the console.  Other well-meaning sysadmins would do this to me all the 
time back when I was a Windows NT Server administrator (TSE was still 
pretty new back then, and Citrix was big bucks).  Thus, you must have 
more than one single display engine and more than one "display control" 
line.  I don't know if someone makes something for personal computer 
architectures that will allow the hookup of more than one physical 
keyboard, mouse, and monitor, and that will actually keep the I/O 
sessions separate.  A KVM switch won't do it because that's going the 
other way--one person using multiple computers.  It's still a single 
"channel", if you will, of I/O; you're in this case just switching that 
single channel to/from a different box.

Ethernet today provides the best, most cost-effective way to do multiple 
I/O channels to and from a single box.  In the case of our beloved 
project, that "fat client" is actually a K12LTSP server, serving 
multiple I/O channels via Ethernet.  The multiple display channels are 
done by TCP/IP and X11, all nice and seamless, for us, exactly so that 
we don't have to go out and look for other, probably expensive, hardware 
to do that.  The hardware to support X11 is cheap; practically any video 
board on the remote terminal will do.  Same for Fast Ethernet and TCP/IP.

Let's now turn our attention back to the LTSP client.  In the case of 
trying multiple I/O channels on a LTSP client, then you'd likely have to 
have a special kernel and other code that understood the hardware 
required to support multiple keyboards/videocards/mice, and you'd have 
to do a bunch of other coding to 1.) keep those I/O channels separate, 
and 2.) correctly send each of them to the LTSP server.  That sounds 
like something for maybe a honors BS Computer Science project or perhaps 
a Master's thesis.

--TP

Steven Santos wrote:

>I was reading /. today (I know, I know - bad idea...), and came across an 
>article on using 1 linux pc for 4 *local* terminals, that is 4 monitors/
>keyboards/mice attached to one local PC, allowing 4 students to use it at 
>once.  I thought it was a cool idea, then my I found myself asking the 
>next logical question: could this be done in an LTSP setup where we could
>effectively eliminate 3 PC's for ever 4 terminals?  Would it save time/$$
>in the long run?
>
>Thoughts?
>  
>