- From: Damodaran, Suresh <Suresh_Damodaran@stercomm.com>
- Date: Fri, 8 Mar 2002 19:17:36 -0600
- To: www-ws-arch@w3.org
Please consider this proposal. Thanks, -Suresh D-AG0007 Proposal ------------------------------------------------------------------------- [1] states D-AG0007 as: "To develop a standard reference architecture for web services that is reliable, and stable, and whose evolution is predictable over time" The goal of this proposal is to identify some "critical success factors" that will help in evaluating these attributes in the emerging architecture, as well as define the requirements the architecture must meet to attain these attributes. This document does not address the critical success factors, though it does address the requirements. I would argue both are the same in this case. General Analysis of D-AG0007 statement -------------------------------------- The goal is to ensure the architecture is reliable, stable, and evolvable. The goal is NOT to ensure web services themselves are reliable, stable, and evolvable. I find this distinction a bit amusing, in that we could potentially have a very reliable, stable, and evolvable architecture that does not ensure the same things for the web services built with it! I will stick to the stated goal of ensuring the architecture has the stated goal, and visit the non-goal if there is public outcry(personally, I think the non-goal is more useful). Reliability of Architecture --------------------------- Note on glossary: Since the WS Arch. would be a framework of standards, I refer to the components of architecture as standards. Reliability is measured against some deliberate/non-deliberate abuses on the architecture. Let me recount such potential abuses: RA1. Implement the interface specified in a standard incorrectly, and complain it doesn't work with another standard within the architecture. RA2. Extend a standard so that it does not work with another standard. RA3. Specify a standard using/depending on at least one proprietary standard. To ensure that the architecture is reliable it should demonstrate the following. R1. All standards should be versioned. R2 Each release of the WS-A also should be versioned with the versions of the standards that go in the WS-A. Such a consistent set of versions will be called a "C-set" - for consistent set. R3. Each such combination, as in R2 must be tested with a set of conformance tests. R4. Extension to a standard by anybody must be submitted to WS-A with a C-set and a set of conformance tests. R5. Non-circular standards: Each standard in the architecture should be specifiable without the use of another standard within the framework. R6. Each standard or the architecture itself should be dependent on non-proprietary standards - standards developed within open standards bodies (W3C, ISO, OASIS,...), where anybody may join and participate in the standardization process by paying the requisite dues. (In the absence of R6, even a single vendor could make WS-A unreliable). Stability of Architecture ------------------------- Stability means a reliable C-set (see R2) does not change arbitrarily. To ensure that the architecture is stable, it should demonstrate the following. S1. WS-A will limit a new release of C-set to once in two quarters, unless serious bugs appear. S2. Only WS-A has the right to release new C-sets Predictable Evolution of Architecture ------------------------------------- I am not sure we can predict the exact way an architecture may evolve. Steps may be taken to ensure that the evolution is along understandable guidelines or axes, though. To ensure the architecture is predictable evolvable, it must demonstrate the following. PE1. The architecture has identified axes for evolution of the architecture. For WS-A, a suggestion of axes are the following [2]. unique identification, - URI/2nd order IDs/... independent specification, - WSDL/vocabulary/ontology interaction - XMLP/intermediary/collaboration/patterned communities PE2. Each standard must be mappable to one of these axes. (I am inclined not to include a standard within the architecture that is mappable to only two or more of these axes, because, then I almost tend call it an non-core standard, outside the scope of WS-A) PE3. The definitions of the architectural elements should be, as far as possible, devoid of jargons and technical terms that may be irrelevant in 10 years PE4. Extension guidelines should be specified for each standard. References [1] http://www.w3.org/2002/ws/arch/2/wd-wsawg-reqs-03042002.html [2] http://lists.w3.org/Archives/Public/www-ws-arch/2002Mar/0113.html
Received on Friday, 8 March 2002 20:18:02 UTC