Version 0.4

[cappelaere]

Just uploaded version 0.4.
I added the draft Workflow Management section (posting new workflows
and definitions) & the search using OpenSearch standard. Other misc
changes based on received feedback.
Let me know what you think.
I appreciate it.
Thanks,
Pat.

[Raj Singh, OGC]

Congratulations to the authors on putting together an interesting
proposal. Here are some comments.

- in the Background section it sounds like you're saying that the WfMC
work is SOAP-based, but this document is all about REST. I think you
mean to say that this document takes the concepts behind the WfMC work
and expresses it in a REST architecture. If that's true, this point
should be more explicit.

- your example in Section 5.2.1 includes username and password as
inputs. Don't you expect most implementations to handle authentication
outside of the specific process? If so, it's misleading to put these
inputs into one of the main examples in the document

- Also in the Section 5.2.1 example you create a resource with an
atom:id of http://geobpms.geobliki.com/wfcs/processes/51.atom. There's
also a <g:wfid> of 20080111-kidawobafu. It seems like the resource is
managed via the atom:id. What's the g:wfid for?

- Based on the ambitious scope suggested by the Workflow Reference
Model Diagram in the Background section, I expected to see discussion
of process chaining within a single Enactment Service, and across
heterogeneous Enactment Services. I also wanted to get some insight
into how much the internals of processes would be described. Did I
miss something?

[Pat Cappelaere]

Raj,

Thank you for taking the time and give us some feedback.
We obviously have some more work to do and I have to admit that this has
been much harder than anticipated.
Other comments inline.

> From: "Raj Singh, OGC" <rajr...@gmail.com>
> Reply-To: <wf...@googlegroups.com>
> Date: Tue, 5 Feb 2008 19:41:05 -0800 (PST)
> To: WfXML <wf...@googlegroups.com>
> Subject: Re: Version 0.4
>
>
> Congratulations to the authors on putting together an interesting
> proposal. Here are some comments.
>
> - in the Background section it sounds like you're saying that the WfMC
> work is SOAP-based, but this document is all about REST. I think you
> mean to say that this document takes the concepts behind the WfMC work
> and expresses it in a REST architecture. If that's true, this point
> should be more explicit.

We are leveraging Wf-XML 2.0 which is SOAP-based. For the record, the WfMC
is also involved in other specifications such as XPDL 2.0.

We do mean to say that WfXML-R is the RESTFul version of WfXML 2.0.
I can try to make this more explicit.

> - your example in Section 5.2.1 includes username and password as
> inputs. Don't you expect most implementations to handle authentication
> outside of the specific process? If so, it's misleading to put these
> inputs into one of the main examples in the document

These are inputs to that particular workflow. This workflow requires to
login to Twitter with a name and password. I am showing later how to
delegate user authority in appendix but this is not required in that
interface.

> - Also in the Section 5.2.1 example you create a resource with an
> atom:id of http://geobpms.geobliki.com/wfcs/processes/51.atom. There's
> also a <g:wfid> of 20080111-kidawobafu. It seems like the resource is
> managed via the atom:id. What's the g:wfid for?

Yes. Actually the atom:id is derived from the resource id.
The g:wfid is really engine implementation dependant and is returned in my
current implementation. I am not sure how other implementations will do.

> - Based on the ambitious scope suggested by the Workflow Reference
> Model Diagram in the Background section, I expected to see discussion
> of process chaining within a single Enactment Service, and across
> heterogeneous Enactment Services. I also wanted to get some insight
> into how much the internals of processes would be described. Did I
> miss something?
>

I am not sure what you expected. Could you expand on this?
Thanks again.

Pat.

[Raj Singh, OGC]

Well, your digram implies that this interface should be able to
facilitate interoperability between multiple workflow engines and
multiple applications. Your examples show a single process invocation.
I want to see examples of a chain of process invocations for a single
workflow engine (sort of like BPEL for BPEl engines), then maybe even
examples of a chain of process invocations for multiple workflow
engines.

--Raj

[Pat Cappelaere]

Raj,

You are correct.

We are actually doing just that for OWS-5 so you will have a chance to see
it in action twice at the end of the month: A workflow running within one
engine instantiating another workflow running on another engine
(OpenWFE-BPEL and OpenWFE-SensorML).

It is using the exact same interface as a remote client would use to
instantiate a workflow. It is a POST to create the workflow resource. In
my case, I created a specific participant that knows how to perform that
activity given the proper input parameters.

I guess that I would need to check with Keith to make sure that the
interface 4 is fully covered in that document.

BTW, I need to add asynchronous transactions... I got it but have not had a
chance yet to update the document... You are keeping me busy :)

Pat.

> From: "Raj Singh, OGC" <rajr...@gmail.com>
> Reply-To: <wf...@googlegroups.com>
> Date: Tue, 11 Mar 2008 13:56:45 -0700 (PDT)
> To: WfXML <wf...@googlegroups.com>
> Subject: Re: Version 0.4
>
>

[Raj Singh]

Great. Can't wait to see it!

--
---
Raj