RE: BA Scenarios: UnsolicitedComplete during CoordinatorCompletion

[Ram Jeyaraman <Ram.Jeyaraman@microsoft.com>]

Thanks Alastair, I have updated the scenarios doc.

-----Original Message-----
From: Alastair Green [mailto:alastair.green@choreology.com]
Sent: Friday, September 01, 2006 5:53 AM
To: Ram Jeyaraman
Cc: ws-tx@lists.oasis-open.org
Subject: Re: BA Scenarios: UnsolicitedComplete during CoordinatorCompletion

Yes, we're forcing a protocol error. We don't have any tests so far that show the reaction of an implementation to a non-conformant interlocutor.

I had something in mind like the following:

Unaltered PS implementation registers for participant completion (creating a state machine that will permit spontaneous Completed).

IA filter (that currently traps messages and then releases them) traps the Register, alters the message to say it's coordinator completion, and delivers the registration.

CS creates a state machine that is CC, and sends RegisterResponse as normal.

CS thinks it's doing CC, and PS thinks it's doing PC.

PC spontaneously sends Completed.

CS gets unsolicited Completed and sends Invalid Protocol? Invalid State?
State, I think.

PS receives unexpected Invalid X, and reports that fact. .

This permits the CS to see what looks like a bugged PS, even though both ends are behaving normally.

Success criteria are a) CS receives Completed in Active, and b) PS receives Invalid State in Completing.

Alastair

Ram Jeyaraman wrote:
> Alastair,
>
> Regarding the unsolicited completed (negative) test case for CoordinatorCompletion protocol:
>
> Do you intend the participant to send forth an unsolicited Complete message (before it receives the Complete from the Coordinator)?
>
> I suppose it may be unreasonable to force the participant implementation (PS) to send an unsolicited Complete message.  Do you mind suggesting a test strategy?
>
> Thanks.
>
>