Re: [Openflow-dev] Re: PROforma papers...

[Bud P. B. <bu...@si...>]

Hi Andrew,

Thanks for the pointer to the overview paper on PROforma--definitely the 
best of all I've seen.

At 01:24 PM 14-03-01 -0800, Andrew po-jung Ho wrote:
>In summary, PROforma looks promising and not overly complex. It could be a 
>good standard for Openflow and OIO to consider adopting. Since I don't 
>have much experience with workflow systems, I look forward to you, Paolo, 
>and other list member's guidance.


In the following are some thoughts on PROforma and workflow models in 
general.  They are based on a still incomplete understanding of the 
situation, but maybe it can help bring the discussion further.

It seems to me that PROforma is highly specialized and far from a general 
purpose workflow model.  Also, apart from workflow, it deals a lot with 
decision support.  Something that general purpose workflow systems usually 
have but that I didn't find in PROforma (did I miss it?) is a description 
of how work flows withing a group of people.  A typical example in an 
administrative organization would be a document/application/etc. to flow 
between people of different roles, each of whom add their contribution...

PROforma seems so specialized to clinical decision support that I wonder 
whether it would be applicable as a workflow component in a LIMS.

So what workflow model/component to use???  A difficult question.  I looked 
around quite a bit so far and haven't found a single thing that convinces 
me all the way (obviously for my purposes).

On the standard side, I'm not convinced that the WfMC or OMG standards are 
really the way to go  (look at the critique by the former IETF SWAP leader 
that I sent you earlier--I don't think these things have really been 
addressed by WfMG's new wf-XML that does not seem to break with the 
problematic reference model).

I spent quite some time reading agent-technology workflow papers and non of 
these people seems to even look at the WfMG standards.  It seems they use 
petri-nets and similar as models and some use some other process 
description standards (that probably are mostly relevant to manufacturing).

So unless there is a strong need for interoperation with a WfMG standard 
complient external workflow system, I'm not sure wether it's worth the 
effort of adding the effort and complexity that comes with the standard.

Other features that are not standardized seem much more important to 
me.  For example, the possibility to negotiate as it is typical for 
agent-based approaches. (As mentioned in section 8.2 of 
http://www.acl.icnet.uk/PUBLICATIONS/ms364.pdf that you pointed 
out).  Also, a peer-to-peer architecture that allows collaboration and 
scalability way beyond the boundaries of a singe administrative 
control.  Such features seems to be almost incompatible with the current 
standards.

Going to actually available systems, my impression is that there are 
roughly three categories of systems that call themselves "workflow" systems:

1. Task-Management systems such as those used in Ticket tracking and 
help-desk systems.  Their basic entity is a task.  The notion of "flow" 
comes in through the possibility of defining task dependencies.  But this 
is very cumbersome and way too limited for many kinds of workflow needs.

Typically, these systems assign human resources to tasks (who is 
responsible for execution, who created the task), but lack any notion of 
other resources, such as for example documents that "flow" around in the 
work process.  (To be fair, some deal with descriptions of the task coming 
via e-mail etc.--but that's very specialized and not general at all).

2. Workflow systems that model an actual flow of tasks by connecting tasks 
in some kind of a graph.  Most serious workflow systems and the WfMG model 
fall into this category.

There is probably a major difference of how systems deal with resources 
that are relevant to work processes.  All systems probably model human 
resources.  Most systems (but not all) allow you that tasks reference their 
input and output resources.  IMO, the mechanism of referencing resources 
from task makes it difficult to understand the flow of resources 
(documents, data) in the workflow.  I personally believe that models that 
emphasize tasks at the cost of resources can be rather limited in some 
kinds of workflow applications (such as LIMS) where resources (eg. samples 
and specimen in a LIMS) are first class citizens.

3. There seems to be a few workflow systems that model workflow as complete 
graphs and also treat resources as first class citizens.  This was my naive 
idea (before I looked into workflow systems in more detail) of a workflow 
component for a LIMS (see 
http://www.sistema.it/labinfo/single.html#Relationships between Entities) 
but also seems to be used by systems such as PIPER 
(http://bioinformatics.org/piper/) and Khoros 
(http://www.khoral.com/ideas/technology/cantata.pdf).

Piper actually looks VERY interesting but it seems too early to be used 
(still alpha).  Interesting that Piper does not seem to bother about any 
workflow standard.  I suppose that treating resources as first-class 
citizens may be incompatible with the WfMG approach.  (Interesting for the 
OIO folks is that the upper layers of Piper are written in Python).

Another product that may be able to manage workflows is the Narval software 
agent (http://www.logilab.org/narval/).  <quote>Narval is the acronym of 
"Network Assistant Reasoning with a Validating Agent Language". It is a 
personal network assistant based on artificial intelligence and agent 
technologies. It executes recipes (sequences of actions) to perform tasks. 
It is easy to specify a new action using XML and to implement it using 
Python. Recipes can be built and debugged using a graphical 
interface.  </quote>

It has been used to automate some special workflows (eg the translation of 
Linux Gazette to French 
http://www.linuxgazette.com/issue59/chauvat.html).  It is written in Python 
and seems to be quite user-friendly and easy to program  (as compared to 
some Java agent systems).  I think it may be worth while to look into 
Narval (and a different paradigm of thinking) to see how that would be 
applicable to workflow in healthcare applications.  My gutt feeling is 
quite positive...

Hope these thoughts are interesting and trigger some discussion.

[I'm not subscribed to the open-outcomes-general list--would you kindly 
copy replies either to ope...@li... or bu...@si...?]

cheers
--bud
/-----------------------------------------------------------------
| Bud P. Bruegger, Ph.D.
| Sistema (www.sistema.it)
| Via U. Bassi, 54
| 58100 Grosseto, Italy
| +39-0564-418667 (voice)
| +39-0564-426104 (fax)
\-----------------------------------------------------------------