Re: [wsbpel] Issue 37 - F2F Presentation Draft

[Alex Yiu <alex.yiu@oracle.com>]

Hi all,

I am thinking out loud again ... :-)

I am thinking this issue from a fresh new viewpoint.

Basically, we have (1A) [always fault, if initiated] and (1B) [fault, if 
initiated and different values] semantics for Issue 37.

Actually, we may be able to have a 3-way switch

"must" :- (1A) semantics; same as what Yaron and Satish want for "yes" 
in previous email
"may"  :- (1B) semantics ; this switch value can be used to handle 
multi-start also
"no" :- trivial semantics


If people want to make sure a CS is initialized in one particular place 
in BPEL and nowhere else, then they will use "must"-"no" combination to 
express that.

If people want to have a session-like situation, they would be happier 
to use "may"-"no" combination for its flexiblity.

(For Java Land people, in servlet API, request.getSession() can be 
mapped to the (1B) "may" semantic. It creates a session, if not exist 
and join/reuse the same session, if it already exists.)

The "may" semantics will solve the mult-start situation also. Two start 
activities are using "may" initiate attribute on the same CS.



What do you guys think?

Again, I am thinking out loud. Please don't blame me too much, if this 
idea does not stick to the wall of TC. :-)


Thanks!

Regards,
Alex Yiu