RE: optional messageReference attribute on interface definition elements

Since (as you have pointed out in an earlier message) the vast majority of
expected WSDL document will be in or in-out interfaces, and these MEPs have
a single message going in a particular direction, the working group decided
to allow this attribute to be omitted.  This will simplify a huge number of
WSDL documents and will avoid the boilerplate 'messageReference="A"' text
for millions of operations.

A substantial saving in bytes and typing if nothing else. :-)

Tom Jordahl
Macromedia Server Development

-----Original Message-----
From: [] On
Behalf Of Yaron Goland
Sent: Friday, January 23, 2004 7:13 PM
To: 'Amelia A Lewis'
Subject: RE: optional messageReference attribute on interface definition

Section 2.4.2 of the spec states "If the {message exchange pattern} of the
Interface Operation component has only one message with a given value for
{direction}, then the messageReference attribute information item  is
optional for the XML representation of the Message Reference component with
that {direction}."

In other words, in one specific case, a MEP with only one message going in a
specific direction, it is not necessary to specify the messageReference
attribute value because the role name can be unambiguously derived from the
MEP's definition.

The change I am asking for is therefore purely a syntactic one and has no
semantic affect on the spec what so ever. That change is to make
messageReference mandatory in all cases. That would mean that rather than
inferring the value of the role name in this single exceptional case, it
will always be explicitly stated.

Since my change request explicitly does not request any semantic changes to
the spec I believe it would be inappropriate to tie or otherwise make my
change request dependent on the outcome of other requests that do ask for
semantic changes.



> -----Original Message-----
> From: []On
> Behalf Of Amelia A Lewis
> Sent: Friday, January 23, 2004 8:49 AM
> To:
> Cc:
> Subject: Re: optional messageReference attribute on interface
> definition
> elements
> I tend to agree, but this was a painfully argument, and the
> result was
> in favor of simplification.  Sanjiva, perhaps you could speak to
> whether this is something that ought to be reopened?
> If we reopen it, can we reopen the discussion of single interface per
> service?
> Amy!
> On Jan 22, 2004, at 6:07 PM, Yaron Goland wrote:
> >
> > The spec identifies one particular case in which the
> messageReference
> > attribute is optional on interfaces. While I am all for
> simplification
> > I
> > think in this case the optionality of the messageReference
> attribute
> > is more
> > likely to lead to confusion about when to use it then to
> lead to any
> > real
> > time savings on the part of the WSDL writer. As such I
> would recommend
> > that
> > for both consistent and clarity we make it mandatory in all
> scenarios.
> >
> > 	Thanks,
> >
> > 		Yaron
> >

Received on Monday, 26 January 2004 10:37:16 UTC