Replied: Tue, 26 Sep 2000 21:16:52 -0400
Replied: "MARLON E.. PIERCE/ADMIN/WPASC/CONTR/SYRACUSE UNIVERSITY" <pierceme@asc.hpc.mil>
Sender: pierceme@asc.hpc.mil
Message-ID: <39D1303C.C9B4952D@asc.hpc.mil>
Date: Tue, 26 Sep 2000 19:24:44 -0400
From: "MARLON E.. PIERCE/ADMIN/WPASC/CONTR/SYRACUSE UNIVERSITY" <pierceme@asc.hpc.mil>
MIME-Version: 1.0
To: fox@csit.fsu.edu, cyoun@csit.fsu.edu, shko@csit.fsu.edu
Subject: XSIL update
Content-Type: text/plain; charset=us-ascii
Content-Transfer-Encoding: 7bit

I've spent some time looking over the XSIL distribution and how we could
use it in the generic object wrapping project.  XSIL provides an XML
based scientific markup language, a connection between the XML and java
objects, and graphics capabilities.  See 

	http://www.cacr.caltech.edu/SDA/xsil

and particularly the documentation xsil_spec.pdf, linked on that page.


* The distribution comes with a simple DTD and java packages for
manipulating the XML (org.escience.XSIL) and viewing it
(org.escience.Xlook).

* The XSIL.java program and its children come with a number of
convenient methods for extracting information from the XML document.  We
should use the XSIL DTD whenever possible, even for non-scientific
input, since this saves us from having to write our own methods for
digesting the XML.  

For example, last Friday the files I sent out included ApplDesc.dtd,
ApplDesc.xml, and parseXMLBean.java.  If I used xsil.dtd instead and
defined all the queue information, etc., using <PARAM> tags, I could
greatly simplify the coding of parseXMLBean.java.  I now just have to
extend XSIL instead of writing all the accessor methods myself.

* As for handling scientific data with XSIL, we have to consider cases. 
For Gaussian, the user manual contains 150 pages of user options, so
constructing a Gaussian input file for the user seems to be out of the
question for now.  Instead, the gaussian.xml file might be as simple as
	<XSIL>
		<Param Name="JobName">test000</Param>
		<Stream Name="InputFile" Type="Remote">test000.com</Stream>
	</XSIL>
and we leave the construction of the input file test000.com to the
Chemistry (and Guassian) expert.

On the other hand, simpler input may be completely specified in XSIL. 
So we need something like
                                       	 __ parseIsingBean.java
				        /																							
       xsil.java-->parseXSILBean.java-- |__ parseGaussianBean.java
					|__ parseParssimBean.java
	                                |...
					|
					\__ parseGenericBean.java					
					

where we have a specialized class to handle the input for a particular
code and a Factory method to retrieve the appropriate class.  The
GenericBean would be a catch-all for new applications.  The <XSIL> tag's
Type attribute can be used to specify which child to use for processing.

* We can use the <Table> tag for input for parametric studies.  Simple
example: a user may want to run calculations at temperatures from 0.1 to
3.0 K, incrementing by 0.1 K.  This job can be summarized by the table

	<Table>
		<Column Name="JobName" Type="string"/>	
		<Column Name="Temp"  Type="float" />
		<Stream Delimiter=",">
			"Job1", 0.1
			"Job2", 0.2
			...
		</Stream>
	</Table>

* I was suprised that XSIL only defines input streams, not output
streams.  Perhaps we will want to extend XSIL to do this.

* The XSIL disrtibution comes with a viewer called Xlook, which may be
extended to provide specialized visualization.  It would be interesting
to see if we could tie this into an applet.  In any case, visualization
is going to be application-specific.

Marlon