RE: Definition of an abstract mep

Youenn, do the two approaches have different component models? --Jeff

 

-----Original Message-----
From: FABLET Youenn [mailto:youenn.fablet@crf.canon.fr] 
Sent: Friday, January 24, 2003 6:17 AM
To: www-ws-desc@w3.org
Cc: Jean-Jacques Moreau
Subject: Definition of an abstract mep

 

Following on discussions at the F2F (monday morning minutes) and two
mails I have just read from Jacek & Glen (who seemed to agree on the
definition of a MEP [1] [2]), it seems to me that there is a
mis-understanding behind the words "abstract MEP".
1)
For some, an abstract MEP specification defines the message exchange
pattern (request-response) and identifies the service within the spec.
From the SOAP request-response mep, two related "WSLD abstract MEP"
would then be extracted :
    - an input-output interaction defined via : <interaction mep=
<http://example.com/mep/in-out> "http://example.com/mep/in-out"/> 
    - an output-input interaction defined via : <interaction mep=
<http://example.com/mep/out-in> "http://example.com/mep/out-in"/>

2)
For others (including me), an abstract MEP spec only defines nodes and
the message exchange pattern between these nodes.
Continuing with the SOAP request-response mep, you will abstract only
one "WSLD abstract MEP" (uri = http://example.com/mep/rr) which defines
two nodes. To define the two above interactions (in-out and out-in), one
will refer to the abstract mep and identify the node that the service
will be.
One syntax is to have two attributes, a 'mep' one and a 'node' one.
    - an input-output interaction defined via: <interaction mep=
<http://example.com/mep/rr> "http://example.com/mep/rr"
node="requester"/>
    - an output-input interaction defined via: <interaction mep=
<http://example.com/mep/rr> "http://example.com/mep/rr"
node="responder"/>
Another syntax is to have a unique attribute pointing to the node name
qualified by the uri of the mep. This syntax is quite similar to the
approach one syntax :
    - an input-output interaction defined via: <interaction
node="rr:requester" xmlns:rr= <http://example.com/mep/rr>
"http://example.com/mep/rr"/>
    - an output-input interaction defined via: <interaction
node="rr:receiver" xmlns:rr= <http://example.com/mep/rr>
"http://example.com/mep/rr"/>

AFAICS, the two approachs are equally expressive and reach the same
level of functionnality (the in-out and out-in patterns can be both
described as well as the in and out patterns).
Comparison between the two approaches :
    - approach 1 and 2 are equally simple at the abstract and syntactic
level.
    - approach 2 conforms with the definition of what is a SOAP mep
    - with approach 2, there are less specifications to be written (with
approach 1, a two nodes soap spec will map to two WSDL abstract mep
specs, a three nodes soap mep will map to three specs...)
    - approach 2 allows the relationship between WSDL files to be made
clearer: the interaction type (i.e. 'abstract mep') would not change
when changing from a node (for instance an http client) to another one
(an http server) wsdl. Only the node name would change

I do not clearly see the advantages of approach 1. Does anybody want to
list them ?

    Youenn

[1] :  http://lists.w3.org/Archives/Public/www-ws-desc/2003Jan/0071.html
[2] : http://lists.w3.org/Archives/Public/www-ws-desc/2003Jan/0032.html