- From: Drew McDermott <drew.mcdermott@yale.edu>
- Date: Mon, 4 Oct 2004 02:34:35 +0000 (GMT)
- To: public-sws-ig@w3.org
This is a response to a note by Naveen Srinivasan in another venue,
reposted here with this permission --
> From: Naveen Srinivasan <naveen@cs.cmu.edu>
>
> Based on our telecon discussion here is an example for unorder construct
>
> Let us consider a composite process composed of 10 different smaller
> processes, and each smaller process purchases an item. Let each of the
> smaller process's process model consist of three parts, first
> withholding $200 from the user's credit card, followed by actual
> purchase (for simplicity let us assume that item's cost is less than
> $200) and finally release the remaining of the $200 previously withheld.
>
> Since there are no dependencies (atleast I/O wise) between the smaller
> process, one can choose to model the above composite process as split or
> split-join. Let us assume that the user's credit limit is $500 and the
> total cost of all the 10 articles he is about to purchase is $75. In
> this case both split and split-join will not succeed, because each of
> the 10 process will try to block $200 and some process may fail.
>
> On the other hand an Unorder process (based on Drew's definition) will
> not fail because Unorder construct will not execute all the processes
> concurrently rather executes them in no particular order.
I have always assumed that "unordered" meant that composite processes
could be interleaved. Your interpretation seems to imply that the
unordered combination of process P_1,...,P_k is eqivalent to a random
ordered permutation of the P_i. Have I misunderstood?
-- Drew
--
-- Drew McDermott
Yale CS Department
Received on Monday, 4 October 2004 02:35:07 UTC