1 Although this example of an implementation strategy is based on tools and directory structures available in the UNIX environment, similar strategies are feasible for other operating systems environments.

[2] In this binding mail address equates to node id

[3] Note that it is possible for two workflow engines to be conducting multiple interoperability dialogues with each other, in parallel, within a given time frame.

[4]The initial assumption for early implementations of this specification is that the value of the workflow_name parameter has been provided by a human agent. It may be that later on more sophisticated ways of ascertianing the value of this parameter might be attempted, but the emphasis at this time is to keep things simple.

[5] Note that this is the id of the invoking process. If the invoking process is itself an invoked sub-process, the id used is that of the invoking sub-process, not that of its parent.

6 May not be known by source workflow engine

[7] Where interoperability is inter-organisational, it may not be appropriate to pass user ids across organiastion boundaries.

8 Information repeated for each attribute value to be set.

9 Information repeated for each attribute value set.

[10] A working assumption is that for efficiency reasons it should be possible to obtain sets of workflow relevant data in a single interaction.

11 Information repeated for each attribute value set.

12 Information repeated for each attribute value to be set.