Re: [ROOT] ROOT, Java and remote script execution

From: Suzanne Panacek (spanacek@fnal.gov)
Date: Wed Oct 04 2000 - 18:03:30 MEST


Hi Anton, Matt, and interested others,

Anton writes:
> >I would like to be able to do is
> > to have a client-server scheme which lets a user to write scripts on
> > his/her side (for example in my R-Quant Data Analysis Studio), execute
> > it on a powerful server where the data files are located and then get
> > back resulting ROOT objects (Graphs, Hists or just numbers).

The RNJ, will allow you to execute the script via an applet and leverage the
power of the server, letting you use the server's disk space and CPU.
However, the applet will not get back ROOT objects. The RNJ is an
implementation of TVirtualX . It is unaware of objects. For example when
drawing a histogram, the RNJ is unaware of what it is drawing it only knows
it is drawing a set of lines and characters. This is what we had in mind for
the RNJ's first iteration, if successful, we would expand it and give it
more intelligence (which is not well defined at this time).

Can I ask you what you will do with the objects?

> >Have you got any farther success with the development?
No, I went to CERN School and am still working on the ROOT User's Guide.

> >Could you comment the above? Could you suggest
> > any existing scheme (DCOM, Corba or whatsoever) that ROOT can try to
> > follow?

An interesting development in this area is Tony's new work on reading ROOT
files in JAS. I believe he instantiates a generic ROOT object in Java, which
then can be used by the Java programmer. This technique, I believe has large
potential to get the objects to the client, and could maybe be used in the
RNJ or in an alternate client/server application of ROOT. Had anyone had a
chance to look at this a little closer?

CORBA/RMI is a good (and official) candidate to communicate objects. Its
drawback is the large overhead, meaning a large development effort, and more
tools to integrate (i.e. IDL compiler).

We could also design our own parallel Java classes that would also be labor
intensive and almost impossible to maintain (who can keep step with Rene and
Fons?).  We could limit the number of classes we implement since now the
dependencies between the ROOT libraries are clearly identified.

> >I assume you
> > have an experience of working with Java and other object distributed
> > client/server models.

I have experience with Java (mostly writing GUI's and a little middle ware),
and I wish I had more experience with client/server models.

Cheers,
Suz.

----- Original Message -----
From: "Matthew D. Langston" <langston@SLAC.stanford.edu>
To: <KOSU_FOKIN@garbo.lucas.lu.se>; <roottalk@pcroot.cern.ch>
Sent: Wednesday, October 04, 2000 9:28 AM
Subject: Re: [ROOT] ROOT, Java and remote script execution


> Tony Johnson at SLAC has written Java Analysis Studio (aka JAS, see
> http://www-sldnt.slac.stanford.edu/jas/).  It is a competing product to
> ROOT, and was designed to implicitly support remote script execution.  It
> has most, if not all (or more) of ROOT's capabilities as far as data
> analysis is concerned.  I believe JAS supports ROOT file I/O now (the web
> page lists all of the features).  Sounds like it may be exactly what you
> want.
>
> Cheers, Matt
>
> ----- Original Message -----
> From: <KOSU_FOKIN@garbo.lucas.lu.se>
> To: <roottalk@pcroot.cern.ch>
> Sent: Wednesday, October 04, 2000 5:51 AM
> Subject: [ROOT] ROOT, Java and remote script execution
>
>
> > Hi Suzanne and rooters,
> >
> > turning back to ROOT/Java bridge... Have you got any farther success
> > with the development? Anyhow, could you comment my thoughts about
> > remote ROOT script execution? It is very good that we can read ROOT
> > files via rootd or httpd but what I would like to be able to do is
> > to have a client-server scheme which lets a user to write scripts on
> > his/her side (for example in my R-Quant Data Analysis Studio), execute
> > it on a powerful server where the data files are located and then get
> > back resulting ROOT objects (Graphs, Hists or just numbers). I assume
you
> > have an experience of working with Java and other object distributed
> > client/server models. Could you comment the above? Could you suggest
> > any existinf scheme (DCOM, Corba or whatsoever) that ROOT can try to
> follow?
> > I think this feature can be very helpful for the whole community.
> >
> > Best regards,
> > Anton
> >
> >
> > -------
> >
> > I also want to reiterate that the ROOT Java package will NOT make the
ROOT
> > classes available in Java. It simply adds the cabability to ROOT to be
> > operated over the net with an applet or application in a client server
> mode.
> > The fact that it is written in Java is acctually irrelevant.
> >
> > Are you planning on using it and if so what for? This is important to
know
> > for me, so that I may consider your needs. Please send me a brief
> > description.
> >
>
>
>



This archive was generated by hypermail 2b29 : Tue Jan 02 2001 - 11:50:34 MET