myGrid Strawman
===============

Chris Greenhalgh
20 December 2002


Work in progress...!!


Experimenting with
 
 * possible port types
 
 * RDF-based metadata 

 * Browser interfaces

for myGrid (http://www.mygrid.org.uk).


To run...
=========

You need HP Labs JENA RDF library; I'm using 1.6.0 at the moment.

Set up the classpath for JENA. Change to the directory containing 
strawman, repository and annotation directories.

Compile the Java files

   javac strawman\browser\CollectionSubtree.java

Run the test application with:

   java strawman.browser.CollectionSubtree repository annotation


Explore...
==========

Source:

	strawman/datatypes - data types :-)
	strawman/ontologies - property & resource constants
	strawman/porttypes - proposed/in progress porttypes
	strawman/repository - metadata and data repositories
	strawman/brower - graphical browser


A few comments...
=================

Initial port types (strawman.porttypes) for: 
--------------------------------------------

 * Data - get value, cf LSID

 * MetaData - simple queries, started like LSID, but is growing
as I explore different kinds of queries

 * MutableMetaData - add/remove metadata

 * Collection - at the moment just a list of things, but has
a place holder for dynamic subset creation c.f. views, or BQS queries.
Waiting on having some interesting metadata to use to make views.

 * Repository - add a new data item.

 * Authority - version of LSID resolution authority; not fully
plumbed in at present.


Initial services (strawman.repository) (all running in process):
----------------------------------------------------------------

 * SimpleMetaDataService - for first and third party metadata. 
Implements MetaData and MutableMetaData.

 * SimpleRepository (extends SimpleMetaDataService) - adds file
store based item storage to metadata, to add implementation for
Collection and Repository, and for its items Data and MetaData. 


The browser (strawman.browser) 
------------------------------

This is based on a tree view (TreeView). Subtrees of this are created by
various helpers:

 * ThingSubtree - creates a node for a named thing, and creates
port-type-specific subtrees for its supported interfaces.

 * DataSubtree - supports retrieval of value from the Data port,
passed to a window based on a HtmlPane (see below).

 * MetaDataSubtree - supports the standard LSID queries on the
item's own metadata interface, displayed on a window based on a
HtmlPane.

 * CollectionSubtree - currently supports listing of contents
on request, generating a ThingSubtree (and so on) for each.

 * RepositorySubtree - offers a file chooser and then a 
very simple (placeholder) window to specify initial metadata 
which is then added via the Repository port type.

The HtmlPane is a demo placeholder which shows how annotations
etc can be added to displayed output in (more or less) the manner
of COHSE. This version looks for LSID URIs and adds links to them
which trigger a (contextual) TextPopupMenu...

A TextPopupMenu is available (a) by right clicking Things in the
browser tree or (b) clicking the inserted links (currently against
LSIDs only) in any HtmlPane. At present this offers links either
(a) an AddCommentWindow for that thing - to add a third party
annotation as a rdfs:comment or (b) a ViewAnnotationWindow, which
uses a HtmlPane to display all known simple statement involving
that URI.

The ViewAnnotationWindow (and the AddCommentWindow) use the
AnnotationManager - which also has a status window - to find
known (currently hard-coded) third-party metadata services that 
they can use.


Next steps/observations...

 * The HtmlPane is a place holder for more general dynamic text
monitoring, markup, annotation, etc.

 * The TextPopupMenu (or equivalent) should also give access to actual
interface functionality for the item (if it is resolvable, e.g. an
LSID); at the moment you can only talk *about* the item, no *to* it.

 * We will definitely need more complex metadata; I've only use simple
triples so far with no intermediate nodes. The APIs and queries will
need extending accordingly.

 * Doesn't have any support for loading ontology stuff or any sort of
reasoning at the moment.


The point of this is rapid exploration, minimal overhead; at augmented
thought experiment.

The route to 'production' or general myGrid code is 
undefined (and may be 'learn from...').


Cheers,
Chris.
cmg@cs.nott.ac.uk

EOF