RE: Editorial issue - Terminology for Operation types

I agree with this editorial change and
with its rationale. I think additional
reasons for making such a change had been
previously suggested in a message from Sanjiva.

One other question I have is whether the WSDL 1.1
terminology--"One-way", "Request-response", 
"Solicit-response" and "Notification"--
can be removed or at least deprecated.

The "Request-Response" pattern can
be variously realized: as a pair
of Inputs, (polarity switch) as
a pair of Outputs, or as specified
as an Input-Output. It is contentious
to pretend that one interface type
is "the" Request-Response pattern.
It also tends to presume that
a "synchronous" communication choice is
essential to the RequestResponse
pattern, and it is not.

As far as WSDL 
entering into the MEP business,
I think it is mixing layers,
and that distinct specifications
from other W3C workgroups should be
created to focus on those layers
of description. Given the rate
of movement here, it might be
pragmatic to divide and conquer.

[Another 400 messages to read.]


-----Original Message-----
From: David Booth [mailto:dbooth@w3.org]
Sent: Monday, May 20, 2002 9:53 AM
To: www-ws-desc@w3.org
Subject: Editorial issue - Terminology for Operation types


This is an editorial issue regarding the terms that we use for the four 
kinds of operations or message exchange patterns described in WSDL: 
"One-way", "Request-response", "Solicit-response" and "Notification".

These four terms are not very consistently selected, nor are they
clearly 
descriptive of their purpose. For example:

1. The word "response" is overloaded.  From the perspective of a Web 
Service that is interacting with a Client, we have the following terms
defined:

"Request-response" represents an input-output pair of messages
"Solicit-response" represents an output-input pair of messages

These terms use different words for the initiating message, depending on

whether the respondent is the Client or the Web Service: "Request" if
the 
initiator is the Client; "Solicit" if the initiator is the Web 
Service.  However, they use the same word ("response") for the returned 
message, even though the  respondent differs.  It would be editorially 
clearer to be consistent.

2. As an operation type, the term "One-way" is not very clear.  It does 
describe a message exchange pattern involving only one message that is
sent 
in one direction (from Client to Web Service).  But the "Notification" 
operation also describes a message exchange pattern involving only one 
message that is sent in one direction -- in this case from Web Service
to 
Client.  Again it would be editorially better to use terms that are more

consistent and/or more distinctly descriptive.

                                  -----

As a solution to these problems, I would suggest using the terms
"Input", 
"Input-Output", "Output-Input", and "Output" instead of the terms 
"One-way", "Request-response", "Solicit-response" and "Notification".
In 
other words, define the following four terms for the four kinds of
operations:

Input (a/k/a "One-way"). The endpoint receives a message.
Input-Output (a/k/a "Request-response"). The endpoint receives a
message, 
and sends a related message.
Output-Input (a/k/a "Solicit-response"). The endpoint sends a message,
and 
receives a related message.
Output (a/k/a "Notification"). The endpoint sends a message.

I think this change would help make the specification simpler, clearer
and 
more consistent.


-- 
David Booth
W3C Fellow / Hewlett-Packard
Telephone: +1.617.253.1273

Received on Thursday, 30 May 2002 12:53:27 UTC