RE: multiple datatypes on a single bound node

Aaron,
We discussed this issue in this week's XForms Working Group
teleconference, and I am reporting the group's consensus and surrounding
discussion.

As you suspected, the results are up to the implementation.  We decided
that there are no errata necessary for this response.

There was some sentiment that precedence order is the wrong pattern to
follow, and open combination may be more appropriate.  Thus, an XForms
processors might choose to satisfy multiple type constraints with a
particular presentation chosen by the combination, and the working group
decided to leave open the freedom to do that.

The WG invites XForms processor implementers and users to respond to
this message with use cases and current implementation practice, in
order to stimulate discussion for possible future enhancements.

Thank you,
Leigh.

-----Original Message-----
From: www-forms-request@w3.org [mailto:www-forms-request@w3.org] On
Behalf Of Aaron Reed
Sent: Wednesday, August 23, 2006 2:57 PM
To: www-forms@w3.org
Subject: multiple datatypes on a single bound node


Hi all,

I saw a discussion thread back in April about whether there is a 
precedence order for a situation when a bound instance node has more 
than one type applied to it (schema type, a xsi:type and a type MIP). 
And I assume that the errata (#10) in the second edition of the spec 
arose out of that discussion, so now we know how to handle validation in

that situation.

But one point was raised in the discussion thread that I didn't see a 
clear answer for.  Consider this scenario...I have a xsi:type="integer" 
on an instance data node, a type="xsd:date" MIP applying to the same 
node and a xf:input is bound to that node.  According to the errata, I 
know to mark the node as invalid since it doesn't validate according to 
all of the types applied to the node.  But I don't know what control to 
bind to the instance data.  Do I bind my control based off of the 
integer type or the date type?

I can see where the WG (and thus the spec) might want to stay mum on 
this subject since binding a particular control based on type is 
implementation specific, but I was wondering if any consensus has been 
reached by implementors?  This is something that I'd think we'd want to 
agree on since instance replacement, for example, could have a 
dramatically different visual result on two different processors if one 
went one way and one goes the other.

Any thoughts appreciated.

Thanks,
--Aaron

Received on Thursday, 7 September 2006 20:46:26 UTC