I'm having some fun with R on a Windows 2003 Server talking to a Wyse
Winterm running Thinstation Linux. The Winterm boots Linux from the
network and then runs rdesktop to a Dell 1750 server (dual 3G Xeon or
somesuch).
The first problem I noticed was that R (and the terminal) ground to a
near halt during demo(graphics). Further investigation showed that it
seemed to be due to having par(ask=TRUE) set. So:
par(ask=TRUE);plot(1:10);plot(1:10)
would cause the Windows session to become unresponsive. Usually it would
recover, but sometimes not.
This was odd, because I have had a similar setup servicing a lab for
over a year without problems. Similar, except this old lab has R 1.9.1
installed and not shiny new 2.2.1. So I installed R 1.9.1 on the new
server - and then there's no problem with par(ask=TRUE) stuff.
Further R installs show the problem manifests itself in 2.1.1 and not
2.0.1, coincidental with a change in the prompt from "Hit <Return> to
see next plot" (2.0.1) to "Click or hit ENTER for next page"
(2.1.1).
Curious...
Running rdesktop from my desktop Linux box and connecting to the same
server doesn't show any of these problems. So I can't blame the network.
Possible reasons for the change:
* Difference in rdesktop versions (1.4.1 on Wyse, 1.3 on my machine)
* Difference in X version (Xorg 6.9.0 on Wyse, XFree86 4.2.1 on mine)
I think the next thing to try is to put R 2.2.1 on the old lab and see
how that copes. Has anyone else any experience running R from Win2003
servers or on Wyse Winterm X-terminals or using Thinstation Linux?
Barry
