Re: [ws-caf] RE: clarification of issue 133

[Greg Pavlik <greg.pavlik@oracle.com>]

I don't think this was fully resolved at the f2f, if memory serves. In 
my opinion, it doesn't make sense for the sender to be arbitrarily able 
to alter context passing semantics, at least as a general rule, which is 
what the current text defines as the default behavior. It seems to me 
that "users" of the context can/should provide the rules for context 
passing. By users, I mean whomever defines the specific protocol. I'm 
not convinced we need to have a special typing mechanism for either 
case, however, if normative requirements will suffice.

Greg

Furniss, Peter wrote:

>I'm sorry - I assumed from the minutes the discussion had been about
>addressing. There is no mention of discussing the fields or structure of
>context, other than the context-manager/content-url elements.
> 
>Was there discussion of the question as to whether a receiver should be
>able to determine by inspection of the held value whether it is a full
>context or should be dereferenced ? If the switching capability is to
>supported (which I have repeatedly said won't work anyway, but everyone
>else seems to think should be allowed), then it is not possible, in
>general, for an implementation to know (even from the context type) what
>it has got.
> 
> 
>Peter
> 
> -----Original Message-----
>From: Mark Little [mailto:mark.little@arjuna.com] 
>Sent: 08 July 2004 09:58
>To: Furniss, Peter; ws-caf
>Subject: Re: [ws-caf] RE: clarification of issue 133
>
>
>
>As I tried to indicate, the decision at New Orleans was to have a single
>context structure and state in the document which fields are required
>for by-reference and which are valid for by-value.
> 
>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: Wednesday, July 07, 2004 2:31 PM
>Subject: [ws-caf] RE: clarification of issue 133
>
>Mark,
> 
>It's not really an addressing matter - though obviously the content of
>the reference form is affected by the NO addressing discussion.
> 
>We seem to agree that (rather obviously) the fields of a
>context-by-value are, in general, different from the fields of a
>context-by-reference (a superset would be a difference). The question
>then is whether this should be expressed only in text or in the schema.
> 
>But first there is a specific question of whether the two forms should
>be explicitly distinguished. Since a full context (i.e. by-value or as
>retrieved) might not have any extension fields or the optional fields,
>it is not possible to tell just by looking at a full context that it is
>in fact full, and de-referencing won't do any good.
> 
>There would seem to be three ways of distinguishing:
> 
>a) there are fields that are always present in one form but never in the
>other
>b) there is an attribute that declares which it is
>c) the structure (xml type) is different.
> 
>The first would work if the context-manager / context-uri choice is
>never present in a full context.  (but if it were, it really ought to be
>a different type). But it would then make it difficult for a
>full-context to be dereferenced again (perhaps to pick up updates that
>have been applied to the master value at the context manager).
> 
>Thus, it seems the effective fields of a full context have to be a
>superset of those in a by-reference. So a) is out.
> 
>That doesn't split b) and c), though c) gives the advantage of forcing a
>by-reference context to contain only the correct fields, or fail
>validation, while b) allows mixtures to be syntactically valid.
> 
>Either way, there is an impact on the schema.
> 
>I would suggest that the fields needed in by-reference are the
>identifier (so the receiver knows if he already has this one), the type
>(so the receiver knows what it is about) and the EPR/uri (so the
>receiver knows where to inflate it).
> 
>Peter
> 
> 
> -----Original Message-----
>From: Mark Little [mailto:mark.little@arjuna.com] 
>Sent: 07 July 2004 11:29
>To: Furniss, Peter; ws-caf
>Subject: clarification of issue 133
>
>
>
>Peter, in order to address issue 133
>(http://services.arjuna.com/wscaf-issues/show_bug.cgi?id=133) I'd like
>to get some clarification. It appears that you are suggesting that we
>have different context structures for pass-by-reference and
>pass-by-value? Is that correct?
> 
>If it is, then I think this issue should be closed because we already
>discussed this at the New Orleans face-to-face in and around the general
>subject of addressing. The result was that the URI that was there has
>been changed to be an EPR, but the structure was to remain the same for
>both formats.
> 
>Following on from the above assumption, we could look at the current
>text and make sure it's clear that only the EPR should be sent if by
>reference.
> 
>Mark.
> 
>----
>Mark Little,
>Chief Architect, Transactions,
>Arjuna Technologies Ltd.
> 
>www.arjuna.com
>
>
>
>  
>