Raw Notes from WFM-RG Session #2 (5:30p on 22 Sep 2004)
-------------------------------------------------------
(condensed by Jim Blythe from original notes by B. Nitzberg)


===== [Jim presented intro slides.]


===== [MiGenAS Workflow Engine presentation presented by Thomas
       Soddemann, RZG]

Q: Explain why application developers can't use BPL
A: Don't want a user to write a BPL workflow description -- user will
make too many mistakes.

Q: How do you do type translation with web services.
A: With a Schema; complex types need to be mapped to Java objects; just
a matter of serializing...


===== Discussion =====

Discussion on the need for a 'simple' workflow language: it should
abstract the details of calling the services so they are not shown to
the user.

WS Choreography is an option for this, however this is aimed at
situations with 2 or more parties and may provide no help for a workflow
with a single point of control. There will be two forms of BPEL: regular
ans abstract, and the abstract form may be what is needed for this.

Some discussion on the relation between choreography and scheduling.

Although BPEL is more complex, it is a standard and so it may not be
worth switching to simpler, but less supported languages. BPEL is an
interchange format, so one could use a higher-level language above it,
e.g. a Java-like syntax or graphical workflow tools.

==== Links with other groups ====

What are the different groups we should maintain links with?
Answers:  JSDL, Life Sciences, GRAAP, CDDLM, Semantic Grid.

JSDL are working on submission information for individual jobs, but are
likely to coordinate with this group on representing the dependencies
between jobs.

There was some discussion on whether we should look at deployment
issues. However, our charter is to consider the whole lifecycle of a
workflow.


====== Next Meetings =====


GGF13 --

  A reasonable number of people are going to Korea based on responses in
  the meeting, however the chairs may not be going. Further discussion
  is needed to determine an appropriate level of involvement at GGF13.

GGF14 --

  Jim's slides include suggestions for workshop topics. There was some
  interest in a workshop at GGF14. Several EU projects in workflow have
  recently started and would be more likely to have results at GGF14
  than at GGF13.

  There are other venues for workshops to consider. There are two
  workshops likely at AI conferences next year. Omer Rana suggests
  CCGrid 05. Deadline for proposals is in November. Those who are
  interested in working on a proposal could email the research group
  list.

======= Group Documents ======

The charter suggests a document on 'workflow specification languages and
tools'. It may be worth splitting this into two documents, one on
languages and one on tools. Something as simple as a set of 1-page
descriptions of tools would be very useful. There is a survey related to
the GGF10 workshop at http://www.extreme.indiana.edu/swf-survey/. A
'living document' that allowed people to submit and maintain
short descriptions of systems would be useful since tools are appearing
rapidly.

It might be worth covering interoperability between different workflow
engines, at different levels, e.g. black box versus full translation
between systems.

It would also be useful to aim for a shared terminology between the
different groups. A reference architecture was also suggested.