RE: Top cloud in triangle/rectangle diagram

At 11:49 AM 10/7/2002 -0400, Sedukhin, Igor wrote:
>Here is my interpretation of the Roger's scenario [at 
>http://lists.w3.org/Archives/Public/www-ws-arch/2002Oct/0072.html]
>
>The guy in Widgets-R-Us who picked up the phone and then received an 
>e-mail with the WSDL did "Find" a service.

You are really stretching the meaning of the word "Find".  The two parties 
already know each other!  There is no "search" whatsoever to "discover" the 
WSDL.

The concept of "finding" or "discovering" an appropriate service is only 
relevant if the two parties DON'T already know who they want to interact with.

>The guy in FredCo who responded to the phone call and then e-mailed the 
>WSDL, played a role of an "Advertizer".
>The other guy in FredCo who created a WSDL or otherwise told the 
>"Advertizer" guy about the WSDL did "Publish" a service.
>
>I think the proper architectural roles were played well :).

Well of course, if you ASSUME that a third role is needed, then you can 
claim that the third role is invisibly played by one of the existing 
parties. But that's an unfair assumption.  My point is that there is no 
need or benefit in hypothesizing this extra, third role in this scenario, 
and therefore we should not.

Consider this Jane/Sue analogy:
Jane and Sue are good friends, living across the river from each 
other.  They decide to build a bridge so that they can easily have lunch 
together every day.  Jane agrees to build her side of the bridge, and Sue 
agrees to build her side.  Jane draws up a blueprint for the bridge, and 
sends it to Sue, who agrees to the design.  They build the bridge from both 
ends, it meets in the middle, and the next day they have lunch at the 
middle of the bridge.

It would be silly to claim that Jane invisibly plays a third party 
"blueprint advertiser" role that enables Sue to "discover" the bridge 
blueprint.  The important thing that allows the bridge to meet in the 
middle is the blueprint itself.  How the blueprint gets from Jane to Sue, 
and how many parties it passes through on its way, is totally irrelevant.

In short, I totally agree that it is POSSIBLE to hypothesize a third party 
role.  My point is that the additional third party role isn't necessary or 
useful to this scenario.  It adds complexity to the architecture that is 
irrelevant to this scenario.

>I hope I'm not trying to be difficult :), but I'd like to see a BASIC WS 
>architecture that does not need the act of meeting two parties and 
>therefore does not need those roles.

Clearly, defining three roles instead of two roles is ADDING complexity -- 
not reducing it.  That additional architectural complexity (i.e., another 
conceptual component) may be needed in SOME scenarios.  That's why it may 
be useful in an EXTENDED architecture.  But it clearly is not relevant in 
this simple, very common scenario.

The important part of this interaction is the WSDL document itself.  THAT 
is what the two parties must agree upon.  (They could send the WSDL by 
donkey caravan for all I care!)  Furthermore, it doesn't matter if the WSDL 
goes through 0, 1, 2 or 20 intermediate parties.  (See the attached slides 
Slide9 - Slide13 for an illustration.) There is no more need to hypothesize 
one intermediary as there is to hypothesize 0, 10 or 20 
intermediaries.  The existence or non-existence of ANY intermediary parties 
is irrelevant to this scenario.


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