Re: "operation name" .. an alternate proposal

Hi Jim,

On Fri, Jul 09, 2004 at 05:55:01AM +0100, Jim Webber wrote:
> Mark:
> > FWIW, I don't even think it's necessary to have the operation 
> > name in the message.  All I request is that there's some 
> > bit(s) in the message that can be mapped via public 
> > specification to the operation name.  For example, I could 
> > "say" via the IETF and IANA, that all messages arriving on 
> > TCP port 15555 are supposed to be stored onto disk.  There, 
> > the operation might be called "store", yet isn't in the 
> > message, but the TCP port is part of the extended envelope of 
> > the message, and therefore is sufficient to identify the operation.
> I might be being dim here, but I don't see how that is possible.
> Messages for different applications will have arbitrary structure and
> content. I accept that within the scope of an application or service the
> contents of a message can be mapped to some logical operation (within
> the scope of the service) but I don't see how a public spec would help
> here.

Hmm, not quite sure I see where the disconnect is.

I wasn't suggesting a *single* public spec handle the needs of a
variety of applications, if that's what you mean.  But the value of
a *public* spec and registration is that anybody receiving a message
should be able to lookup - using only the registry (IANA) and the bits
that form the message - the public spec which describes the semantics
of that message (at least to the point where the spec defers to
another spec, as, for example, HTTP does with the Content-Type header).

> But I do agree that the operation name is implicit not explicit - though
> I maintain it is implicit in the message and resolved by the Web
> Service.

Resolved in what way?  I thought that the operation name was always
"processThis"[1] in your view of this.  No?

> > I think that's the minimally acceptable scenario.  IMO, 
> > receiving a message and not knowing what's being asked of 
> > you, is simply not an option and should be actively discouraged.
> This situation never occurs though does it?

I hope not, and I've never seen it myself, but it seems a natural follow
on from some of the non-self-descriptive suggestions that have been made
in the past, and also seems to be what Jeff was suggesting;

"WSDL 2.0 should not require identifying the operation name because doing
so will unnecessarily limit the applicability of WSDL 2.0."

> I advertise some contract
> that says "Send me a 'WeatherUpdate' message" and since I advertised
> that contract it would be ridiculous of me to then not understand those
> messages.

I think message semantics need to be understood independent of any
advertisement, but yah.


Mark Baker.   Ottawa, Ontario, CANADA.

  Seeking work on large scale application/data integration projects
  and/or the enabling infrastructure for same.

Received on Friday, 9 July 2004 07:31:51 UTC