[Date Prev] | [Thread Prev] | [Thread Next] | [Date Next] -- [Date Index] | [Thread Index] | [List Home]
Subject: RE: [ws-caf] RE: close issue 134?
I presume we can discuss this issue today at the F2F. It is consonant with a number of Peter's outstanding issues. In a previous post I raised the point that the issue of children contexts either makes a must-understand per context necessary, or may obviate it (depending on where children contexts end up: in or out). In general, I am concerned about the approach that says: "we can foresee that some referencing (extending) specifications may need something like this element (although we cannot be sure how they will use it, or what its content will be, or whether all contexts will need it), so therefore we should include it in the base context." I incline to the view that the base context should only contain those elements which must be used by all contexts, and whose meaning is not dependent on a subsequent extension or referencing specification. This also has implications for the meaning of the complete message. Another way of putting this would be: the WS-Context spec should be independently useful, and not be laden down with lots of optional fields that might only come into play with *some* referencing specs. Otherwise we run the risk of trying to half-define a subset of the universe of future referencing specs. This discussion might also be aided by the examination of WS-CF and WS-TXM specs as examples of referencing specs. They tend to "do their own thing", and I think this reflects the reality that their requirements are specific. Another case that bears on this is the status type and completion status type. What if the element used for completion status in a referencing spec is a complex type? Is it right to restrict the base type to a string? And is it clean to end up with the moral equivalent of java.lang.Object, where we have <wrapper> <type/> <contents/> </wrapper> in various places? Wouldn't it be better to allow the referencing spec to define its own signals and their fields in its own, natural way? Another related example on the signals front: in BTP there is a control protocol, which allows differential coordination. If a client issues a PREPARE_INFERIORS or a CANCEL_INFERIORS then the signal PREPARE or CANCEL is transmitted on to a defined sub-set of participants. The same applies to the completion signal CONFIRM_TRANSACTION, which can cause CONFIRM to go to some inferiors, and CANCEL to others. Alastair -----Original Message----- From: Mark Little [mailto:mark.little@arjuna.com] Sent: 13 July 2004 09:44 To: Doug Bunting Cc: Furniss, Peter; ws-caf Subject: Re: [ws-caf] RE: close issue 134? > I am not sure what you mean in your last sentence since it seems to be > missing a verb. However, I believe you have a reasonable summary in your > middle paragraph. My summary might have been a tad more general (avoiding > specific solutions): Put it down to typing dyslexia. Mark. > > "Should the WS-Contest specification (or schema) provide an in-line > mechanism for identifying which added content must be understood?" > > Generalizing the question allows for either a global attribute for use on > any extended content or a local attribute on the container itself (which > might repeat). And, yes Peter, we have a question for the "by reference" > case since it is unclear (a priori) who all might see a particular context > structure and therefore what each "viewer" must understand. > > thanx, > doug > > On 12-Jul-04 08:27, Mark Little wrote: > > Doug, in fact I think that's where the idea of the context mustUnderstand > > originally came from. So let's see if I can summarise this as an issue (and > > feel free to correct/augment): > > > > should the extensibility element have mustUnderstand associated with it? > > > > Or are you mustUnderstand for more elements in the context structure? > > > > Mark. > > > > ----- Original Message ----- > > From: "Doug Bunting" <Doug.Bunting@Sun.COM> > > To: "Furniss, Peter" <Peter.Furniss@choreology.com> > > Cc: "Mark Little" <mark.little@arjuna.com>; "ws-caf" > > <ws-caf@lists.oasis-open.org> > > Sent: Monday, July 12, 2004 4:24 PM > > Subject: Re: [ws-caf] RE: close issue 134? > > > > > > > >>Peter and Mark, > >> > >>Issue 129 did not "remove mustUnderstand entirely" but instead dropped a > >>WS-Context specific description of this SOAP attribute. We deferred any > >>specific semantics or requirements for this attribute to the referencing > >>specification, if that proves necessary. > >> > >>For issue 134, I believe we still need a ctx:mustUnderstand attribute > >>because soap:mustUnderstand does not address understanding of information > >>extending the base WS-Context structure. We have an extensibility point > >>that may contain information that both sides must understand but may also > >>contain information of interest primarily to one side only. While I can > >>imagine that many referencing specifications would clarify where > >>information might be added that is of interest only to one side (say, > >>internal references you need when processing the set of related messages > >>defined for a context type), I do not think it appropriate to require full > >>a priori knowledge of this important facet. As a general practise, > >>extensibility points should support in-band identification of the content > >>that must be understood. > >> > >>thanx, > >>doug > >> > >>On 06-Jul-04 05:59, Furniss, Peter wrote: > >> > >>>Mark, > >>> > >>>Just to be clear, there are (or were) two mustUnderstand attributes > >>>referred to in 0.3-and-earlier - one defined in SOAP, which the WS-CTX > >>>spec said had to be ="true", and one defined in ws-context schema as an > >>>attribute of participating-services-list, alongside the mustPropagate > >>>attribute. > >>> > >>>129 concerned only the SOAP one. > >>>134 concerned mostly the ws-context:mustUnderstand, and a passing > >>>mention of mustPropagate. The latter was removed by 131. > >>> > >>>I am pleased at this resolution. > >>> > >>>Peter > >>> > >>> -----Original Message----- > >>> *From:* Mark Little [mailto:mark.little@arjuna.com] > >>> *Sent:* 06 July 2004 13:45 > >>> *To:* Furniss, Peter; ws-caf > >>> *Subject:* Re: [ws-caf] RE: close issue 134? > >>> > >>> Yes, 129 removes mustUnderstand entirely and 134 does likewise with > >>> mustPropagate. > >>> > >>> I'll mark the issue as closed and refer to those other issues. > >>> > >>> Thanks, > >>> > >>> Mark. > >>> > >>> ----- Original Message ----- > >>> *From:* Furniss, Peter <mailto:Peter.Furniss@choreology.com> > >>> *To:* Mark Little <mailto:mark.little@arjuna.com> ; ws-caf > >>> <mailto:ws-caf@lists.oasis-open.org> > >>> *Sent:* Tuesday, July 06, 2004 1:40 PM > >>> *Subject:* [ws-caf] RE: close issue 134? > >>> > >>> Yes, if I understand the intent correctly > >>> > >>> 129 concerned the SOAP:mustUnderstand attribute in > >>> ws-context:context elements in SOAP Headers. The resolution of > >>> 129, as I understand it, will be to remove any specific > >>> statement about the SOAP:mustUnderstand (it may note that the > >>> attribute is available and can have either value, as is normally > >>> the case for SOAP Header elements. > >>> > >>> 134 concerned the wsctx:mustUnderstand attribute which was > >>> (incorrectly) defined as an attribute of > >>> participating-services-list. I assume the intent is to remove it > >>> completely. If that is the intent, I concur. > >>> > >>> Peter > >>> > >>> -----Original Message----- > >>> *From:* Mark Little [mailto:mark.little@arjuna.com] > >>> *Sent:* 06 July 2004 12:24 > >>> *To:* Furniss, Peter; ws-caf > >>> *Subject:* close issue 134? > >>> > >>> Peter, I believe that this issue > >>> > > > > (http://services.arjuna.com/wscaf-issues/show_bug.cgi?id=134) > > > >>> is no longer required because of the resolution to issues > >>> 129 and 131. Since you raised it I wanted to check before > >>> doing anything (or proposing to do anything). > >>> > >>> Mark. > >>> > >> > > >
[Date Prev] | [Thread Prev] | [Thread Next] | [Date Next] -- [Date Index] | [Thread Index] | [List Home]