- From: Sergey Beryozkin <sergey.beryozkin@iona.com>
- Date: Fri, 20 Oct 2006 10:32:25 +0100
- To: "Yalcinalp, Umit" <umit.yalcinalp@sap.com>, "Frederick Hirsch" <frederick.hirsch@nokia.com>
- Cc: "William Henry" <william.henry@iona.com>, <public-ws-policy@w3.org>
Hi "Overloading the wsp:optional marker just complicates the matter, because it provides a cop-out for not understanding the semantics of the assertion by creating an alternative that only a class of clients will understand and will engage with." That's what wsp:optional is all about from a requester's perspective, isn't it ? It translates to a normal form expression with two policy alternatives, a class of clients which can not understand one alternative can choose to ignore it and select a diff alternative. "In terms of logging, I do not believe that it should be implemented by optional" +1. It has to be a wsp:local (custom:local) and stripped out out of the WSDL if possible (if custom:local then must be stripped), otherwise ignored by a requester. Put wsp:local on assertion which can not be of interest to requesters, otherwise use normal assertions, doesn't matter optional or not. There's no way one can prevent a provider from misusing wsp:optional by exposing an assertion like <custom:useLocalTimerForServerLogging/>. Hence I believe wsp:optional wording should be improved and simplified, with a clear direction to policy authors. Thanks, Sergey
Received on Friday, 20 October 2006 09:31:24 UTC