JISC Workflows meeting (13/02/07)

The purpose of this meeting was to see if and where there are commonalities between workflows and to see if there are any common points between domain specific workflows.

The agenda was very full with six presentations from six very diverse projects (ISIS/ASSIS; RepoMMan, Human Collaborative workflow; ePHPX(escience);COVARM;Kuali).

Steve Jeyes and Warwick Bailey described there experiences of using IMS Simple Sequencing, QTI and BPEL. They were surprised at how easy it was to use BPEL. This was due partly to the Active Ends visual editor. Warwick did point out that more work needs to be done to clarify just what is valuable about using BPEL. He proposed that it might have something to do with the ease of use and the ability to have long running calls and the use of xpath; but he would like to seem more work done in this area. He also stressed the importance of xsds and how the skill of creating elegant, extensible xsds is really undervalued. At Icoden they have found the .NET toolkit easier to use than java, but he did point out that may just be a personal preference.

Steve Jeyes highlighted the problems his team had with using simple sequencing (or not so simple sequencing as it maybe should be called) and the need for more work to be done in terms of integrating standards and workflows.

Richard Green from RepoMMan project then outlined some of the workflow issues they have been grappling with in their project and within the wider institutional context. The University of Hull’s vision of a repostitory encompasses storage, access, management and preservation of a wide range of file types from concept to completion. A user survey highlighted that their system users (primarily researchers) wanted a safe place which could be accessed anywhere, anytime and had support for versioning. So they have been creating a toolset to manage workflows for users and they have found UML useful for creating basic workflows. They are also trying to add in as much automation to workflows as possible for example by pre-populating metadata fields by using JHOVE (which btw he seemed very excited about as it actually does seem to do a lot of pre-populating of fields) and trying to get as much as possible from other services.

Scott Wilson then looked at issues surrounding human collaborative workflows (the non BPEL stuff :-)). Scott outlined the work he had been doing at McQuarrie University in relation to collaborative research practice and the development of RAMS (research activity management system) from LAMS. They have been looking at learning design as potential workflow method as there hasn’t been a lot of work done around communication and collaborative methods as workflows. One common characteristic of the research process is that the process can change at various stages during the lifecyle and very few systems support the levels of flexiblitity at runtime that this requires ( this is also true of learning design systems). Scott also pointed out the risks of trying to develop these types of systems when compared to the actual benefits and how easy it could be to develop systems for experts rather than practitioners (again very similar to learning design). One of the key issues for this work is the fact that in collaborative settings, seemingly simple workflows can actually exhibit complex behaviour which again reinforces the need for adaptable systems. In Scott’s opinion, collaborative processes don’t lend themselves to todays business process model methods. But they are hoping that the RAMS system will be a step in the right direction.

Rob Allan from the ePHPX then gave an eScience take on workflows. Naturally this sector is very concered with provenance, the use of metadata and authorisation to re-use data. One problem eScience has is that each domain within it tends to invent their own domain tools and he would like to see more work done on creating webservices that could be shared. He emphasised the need to make workflows easy for users and the need for guidelines and tutorials for the creation and use of webservices. There are some example tutorials available @ http://www.grids.ac.uk/WOSE/tutorials.

Rob also highlighted the need for well defined data modesl and/or semantic tools to support data interoperability between applications linked to workflows.

Next up was Balbir Barn talking about the approach taken by the COVARM project. During the project they used a UML model based solution. They used scenarios to identify services and workflows, and these matched quite closely to BPEL process definitions.

The experience of the project has reinforced the team’s belief in the model driven activity approach. He believes that there is a need for better methodology to support eframework activities and their approach could very well fit this gap. The domain information model they produced has a structural model which can be used to identify services. The synthesis of processes can provide a mapping to BPEL. However there are some techincal issues with this approach. Although UML models are mature there are some issues within the soa context. There is a need for testing framework requirements and to be able to see the service view and overall business process view in parallel.

The final presentation of the day was from Barry Walsh and Brian McGough from the Kuali project. The Kuali Enterprise Workflow started with financial information but has now broadened out to integrate student systems, research systems and HR – areas Barry described as ‘human mediated activities’. They had to develop robust and scalable functionality whilst remaing user centric. The ability to allow people to do things ‘in their way’ was fundatmental. They have developed a generic workflow engine which supports any kind of typical process within the areas they are working in.

Unfortunately I had to leave before the discussion session but some of the key messages I got from the day were:
*there isn’t a lot of convergence around workflows – people still want to do it their own way.
* more work needs to be done defining the differences between automated workflows and human workflows
*hand off points need to be clear, and we need to be able to identify appropriate tools/services for these points

A respresentative from the Mellon foundation attend the meeting and as far as I can gather JISC and Mellon are going to continue a dialogue around funding for workflow projects.

2 thoughts on “JISC Workflows meeting (13/02/07)

  1. Hi Sheila, enjoying your blog and finding it highly informative on a regular basis. This piece on workflows is highly pertinent to us at Intrallect as we have a number of customers who use complex workflows within intraLibrary. Will there be a formal report of some kind from this meeting? I’d like to be kept informed of any future work in this area. The info on and link to JHOVE was particularly useful- great service from the JISC CETIS Educational Community Queen! (Is that your current job title? I can’t keep up :-) ).

  2. Hi Sarah

    Glad to hear you are enjoying the blog:-) I’m not sure if there is going to be a report from that meeting, but I’ll try and find out and keep you posted.

    Cheers

    Sheila