[kepler-dev] A very thorny question..

[Glen Jarvis]

Dan,
   Which would be more appropriate;  a private response or a response  
that included a carbon copy to kepler-dev?

   The platform that I tested on was: Mac OS X 10.4.9 (Darwin 8.9.1  
Kernel). The machine contained a 2 GHz Intel Core Duo processor and  
had 1 GB RAM.

   When, I went through the Getting Started Guide a second time,  I  
used the  nightly build for 20070325. It is on this build that the  
errors that you saw were reported and only fixed when installing R  
separately. I used the nightly build for testing instead of the Beta  
for testing because the guide itself clearly was written for more  
recent versions of the software (and new actor names).

   When I initially installed the software (on my initial introduction  
to Kepler), I used the Beta3 install  and went through the Getting  
Started Guide. At that time, I don't remember exactly what worked and  
what didn't. I was only reviewing for my own sake and to get a feel  
for how the software worked. I remember that several examples did not  
work, but I cannot say that the R examples certainly did not. I do  
remember struggling because the actors names had changed. But, using  
the new nightly build obviously fixed that piece of the problem.

    If you include R in all the Beta releases, and if you intend to  
include R in *all* distributions that are released, then almost every  
single concern of mine is resolved. I could understand if R is not  
included in nightly builds but is done so in releases.

    The one last problem that would still exist, regardless of R,  
would be the external browser communication. My belief (although  
untested) is that if you were to install Netscape (not Firefox, but an  
actual Netscape install where the command 'netscape' from a command  
line would start the browser), that suddenly this problem would "go  
away." If this were true, then there was no mistake when testing if  
Netscape happened to be installed on the machine that was used for  
testing.

    The deeper question I am concerned with is what systematic  
approach is being used so that the BrowserDisplay actor will always  
find "the proper browser" for all platforms? It's not an easy question  
to resolve. I hopefully have answered all of your questions except for  
the "File Error" message. Since the demonstration worked other than  
reporting that message, and in light of these other questions, I did  
not investigate further. So, for that question, I am sorry I can't  
answer it immediately.

Warmest Regards,


Glen Jarvis
-- 
913-486-8775
glen at glenjarvis.com
http://www.glenjarvis.com

"You must be the change you wish to see in the world." -M. Gandhi




Hi Glen,

    Thank you for your comments. I agree that all the examples in the  
Getting Started Guide should work! I thought they did, at least for  
the Kepler-beta3 installer version on Windows. [I put together the  
installer.] I apparently made a mistake, because I see a problem with  
the BrowserDisplay actor replacement as you mentioned even on a  
Windows box with Beta3. However, the StatisticSummary example does  
work for me on Windows and the Windows installer does include R (not  
the nightly build, however). So it does apparently run for most  
Windows users (which are the majority of our users.) We may well want  
to include R with Mac and Linux versions in the eventual release. [The  
problem is the size of the installation as we keep adding more code;  
we do consider R to be very important and very useful.]

    What OS are you running? I have not seen your "File Error" message  
on the Mac or Linux when R is installed. Can you provide me further  
information on this?

Dan Higgins
NCEAS - UC Santa Barbara