Hi Rex, Michael, all:
I want to resist the temptation to reply to every comment made against our draft. I have my opinion too but Chris and I have tried to keep those out of the picture, concentrating instead on finding clearer wording and compromise on issues which – when discussions are held and people understand each other’s PoVs – seem in most cases to be pretty much close already.
On the issue of mediator, I did stick by neck out and I’ll explain. I do not want any member of the TC to feel we have simply ignored their point.
There are two key issues for me:
- I believe that there is a formal modelling error in describing ‘a mediator’ (and for that matter ‘a consumer’ or ‘a producer’ also) as ‘a participant’. I have argued that there are not sub-classes of participant but rather roles that participants play in different contexts. The roles of mediator, consumer and producer (possibly, probably, other roles too) can be reified and modelled as needed but – as far as the concepts ‘participant’ and ‘service’ are concerned – they are associative roles.
Therefore: Mediator is not a participant
- Any common sense reading of what a mediator does would point to it being something offered (a service, dare I say) to mediate between two or more others participants’ (potentially) conflicting goals, needs, capabilities, etc. I would argue therefore that mediator is a service and, yes, it is also a role played by a participant. From a modelling point of view it gets a bit complex (but not unmanageable) because a service-that-is-a-mediator may in some contexts also be playing the role of a service consumer and/or service provider. Likewise, services – as they may be set up – could play the roles of mediator, repository, registry, and others but, the roles played are (again) context dependent. BTW, It is one reason that I would argue that using only UML class diagrams to model what we want to capture may not be enough – this is an example where state diagrams, for example, might add further clarity.
We have a couple of weeks to digest the new draft and plenty of comments in the meantime, so I shouldn’t pronounce further today on what wording might be appropriate, but I do want our ontology to stand up to scrutiny – those of you following the Ontology Forum will see the mauling currently being given to the Open Group’s SOA ontology work (rightly, in some ways, I would say). I don’t want us to suffer the same fate.
A final comment: after 30+ years working with politicians, I can definitely state that compromise wording may help smooth ruffled feathers, but it doesn’t mean that the result is any more meaningful or tractable! What we are aiming for is something that makes sense, has broad support and stands up to scrutiny.
Best regards and talk soon,
Peter
From: Rex Brooks [mailto:rex.brooks@ncoic.org]
Sent: Wednesday, 22 December 2010 16:33
To: mpoulin@usa.com
Cc: peter@peterfbrown.com; soa-rm-ra@lists.oasis-open.org
Subject: Re: [soa-rm-ra] Groups - New Editors' Draft, 21 Dec 2010 - Clean (Editors' Draft-2010-12-21 (Clean).pdf) uploaded
Hi Michael,
I don't have an opinion yet on 1, but for 2, I don't think Chris and Peter ever bought into the notion of our discussion as constituting an agreement for them. In any event, we can discuss this in the meeting. I would suggest to the group that what Michael and I agreed on may be a good choice. At the very least it represents a compromise and that, in itself, is a very positive development.
I will write separately on the main concern that I had from my first pass through Section 3.
Cheeers,
Rex
Cheers,
Rex
On 12/22/10 2:48 AM, mpoulin@usa.com wrote:
I have started reading this edition and already have two comments for your attention:
1) We list all SOA related standards from The Open Group: why we do this? It is read as our agreement with them, which is not necessary the fact because we have not verified compatibility between them and RM/RAF. For example, I know that SoaML defines SOA service as service interface. I even wrote an article about this.
2) definition of Mediator: " A service mediator is a participant who is facilitating the discovery, offer or use
789 of services in some way" is not the one we agreed with Rex.The agreed definition was: "A service mediator is participant that facilitates interactions and connectivity in the offering or use of services. "
- Michael