All:

On the DFDL WG call  17NOV2010 my concern was that this rule was added 'recently' but on closer review I see that it was just changed a bit ("discriminators" changed to "dfdl:discriminators").  Ruth, Alan and I are all  in agreement  that this rule is a throwback to an earlier discrimination schema and needs to be removed.

The proposal is to remove the following from the spec:
In section 7.4 : "There can be at most one dfdl:discriminator  within a sequence group. It is a schema definition error if there is more than one dfdl:discriminator  annotation that applies to a single sequence group

I'm sending this email out with two purposes:
    1. Does anyone disagree and believe that this rules is still relevant?
    2. Is this change critical enough to amend the final spec or publish some form of errata to the specification? And if so, what is our next step?

Regards,

Stephanie Fetzer
WebSphere Transformation Extender (WTX)
Industry Packs - Architect

----- Forwarded by Stephanie Fetzer/Charlotte/IBM on 11/18/2010 07:54 AM -----

From:	Alan Powell/UK/IBM@IBMGB
To:	Stephanie Fetzer/Charlotte/IBM@IBMUS
Cc:	Ruth Wiegand/Boca Raton/IBM@IBMUS, Steve Hanson/UK/IBM@IBMGB
Date:	11/17/2010 05:55 AM
Subject:	Re: DFDL Discriminator - Question from Ruth

---

Stephanie

I think you are correct. That statement is probably left over from when the discriminator confirmed the existence of a parent.

Regards

Alan Powell

Development - MQSeries, Message Broker, ESB

IBM Software Group, Application and Integration Middleware Software

-------------------------------------------------------------------------------------------------------------------------------------------

IBM

MP211, Hursley Park

Hursley, SO21 2JN

United Kingdom

Phone: +44-1962-815073

e-mail: alan_powell@uk.ibm.com

From:        Stephanie Fetzer/Charlotte/IBM@IBMUS
To:        Steve Hanson/UK/IBM@IBMGB, Alan Powell/UK/IBM@IBMGB
Cc:        Ruth Wiegand/Boca Raton/IBM@IBMUS
Date:        16/11/2010 20:42
Subject:        DFDL Discriminator - Question from Ruth

---

Steve/Alan:

Was just talking with Ruth about the discriminators and I think we've got something in the spec that we may need to remove in this area.  Let me talk this out here first to see if perhaps I've remembered something incorrectly.

In section 7.4  we have: "There can be at most one dfdl:discriminator  within a sequence group. It is a schema definition error if there is more than one dfdl:discriminator  annotation that applies to a single sequence group"

This makes perfect sense in WTX.  In WTX an identifier is placed on a component of a group. Then the identifier is proven to exist the parent exists.  Allowing only one per the sequence is necessary in this scheme.  
In DFDL we place the discriminator on the component being discriminated. So we have a rule that points to ones children such as ..if present child2 then I exist.

I can not see how the 'one per sequence' gets us anything in CT in most cases.  We can't mean that only one of the parents in a sequence can contain a discriminator.  In EDI each segment in a loop can have a discriminator.  As we are using a rule at the parent it would be difficult (and not very useful) to restrict the rule to referencing only one child.   So I'm confused - and I've confused Ruth.

Can we take five minutes from the WG call to straighten me back out again please?

Regards,

Stephanie Fetzer
WebSphere Transformation Extender (WTX)
Industry Packs - Architect