[Date Prev] | [Thread Prev] | [Thread Next] | [Date Next] -- [Date Index] | [Thread Index] | [List Home]
Subject: RE: [ws-rx] Issue 060 ammendment to current proposal
Although I favor removing them, as an alternative, I would highly recommend that we at least clarify the abstract nature of AD in the model and make a normative statement allowing multiple implementation scenarios. I feel it is too restrictive as written and prone to similar interpretation. At present, we claim this is a concrete specification and the word "abstract" appears in only two places in the specification. Delivery is defined fairly concretely IMO. D ******************************* Adobe Systems, Inc. - http://www.adobe.com Vice Chair - UN/CEFACT http://www.uncefact.org/ Chair - OASIS SOA Reference Model Technical Committee Personal Blog - http://technoracle.blogspot.com/ ******************************* -----Original Message----- From: Anish Karmarkar [mailto:Anish.Karmarkar@oracle.com] Sent: Thursday, December 08, 2005 12:10 PM To: Duane Nickull Cc: Doug Davis; ws-rx@lists.oasis-open.org Subject: Re: [ws-rx] Issue 060 ammendment to current proposal Duane, AS, RMS, RMD, AD are part of the abstract model and does not necessarily reflect the underlying implementations architecture. I would think that your example of the RMD delivering 1st message to "endpoint" 1 and the 2nd message to "endpoint" 2 is quite valid per this model, as this model would consider "endpoint" 1 and 2 as a single entity: AD. The abstract model is based on functionality/responsibility of various parties/entities involved in the protocol. It allows us to define/describe the protocol and specify responsibilities/contracts. -Anish -- Duane Nickull wrote: > Part of me wonders if we should just be talking about RMS and RMD and > not AS/AD at all. > -Doug > > All of me wants to drop AD/AS. > > 1. Implementers should be free to design and control how they process > messages "in order". As long as they understand the intentions of > the RMS (re-create the stream) and can make decisions on that, > that is all we should care about. > 2. The WS-RX spec should not constrain all service implementations to > a specific model or pattern of AS -> RMS -> RMD -> AD > 3. We should not constrain service by mandating that any message or > message sequence received by the RMD cannot be divided to multiple > endpoints. The way AD is worded now, it appears that all messages > of one sequence must be "delivered" to another single actor in > order. What if someone wants to have the first in a sequence go > to endpoint #1, the second go to endpoint #2 etc? The RMS should > not care as long as the "externally visible aspects" of the > service match their expectations. > > There are several other reasons but for these alone I would prefer to > drop AD, AS from the spec and perhaps replace it with some words that > reflect the real intention which I believe is to allow the RMD to > understand the sequence intentions of the RMS yet leave implementers > free to handle the intentions in a way they see fit. We should stick to > "externally visible properties" only and not dictate specific models > behind a service interface. > > Duane >
[Date Prev] | [Thread Prev] | [Thread Next] | [Date Next] -- [Date Index] | [Thread Index] | [List Home]