From: Veerappan Saravanan-ESV007 <esv007@motorola.com>
Date: Wed, 5 May 2004 09:43:33 -0400
To: "'noah_mendelsohn@us.ibm.com'" <noah_mendelsohn@us.ibm.com>, Veerappan Saravanan-ESV007 <esv007@motorola.com>

```
Thanks for the explanation.

I am trying to figure out alternates for the following due to UPA violation as per spec. I need to perform this validation in schema
itself rather during parsing.

<choice>
<sequence>
<element ref="a" maxOccurs="1"/>
<element ref="b" maxOccurs="5"/>
</sequence>
<sequence>
<element ref="a" maxOccurs="1"/>
<element ref="b" maxOccurs="1"/>
</sequence>
<sequence>
<element ref="a" maxOccurs="5"/>
<element ref="b" maxOccurs="1"/>
</sequence>
</choice>

Thanks,
-Saravanan

-----Original Message-----
From: xmlschema-dev-request@w3.org [mailto:xmlschema-dev-request@w3.org] On Behalf Of noah_mendelsohn@us.ibm.com
Sent: Tuesday, May 04, 2004 10:50 PM
To: Veerappan Saravanan-ESV007
Cc: xmlschema-dev@w3.org

Veerappan Sravanan writes:

>> UPA
and how to get around?

I must ask:  "better than what"?  Here's a quick intro:

First, what's a particle.  Consider:

<sequence>
<element ref="a"/>
<element ref="b" maxOccurs="5"/>
<element ref="c">
</sequence>

In the above, there are 4 particles[1].  One is the outer sequence, the
others are the three element references.  The maxOccurs is considered part
of the particle.   Particles have recursive structure:

<sequence>
<element ref="a"/>
<element ref="b" maxOccurs="5"/>
<choice>
<element ref="x"/>
<element ref="y"/>
</choice>
<element ref="c">
</sequence>

Anyway, UPA says that for a given content model to be legal, there can be
no instance that would have an element matching more than one particle.
The above are OK per UPA.  The following is not:

<sequence>
<element ref="a" minOccurs="0"/>
<element ref="a" minOccurs="0"/>
</sequence>

Why illegal?  Well, if I asked you to validate the instance:

<a/>

you could match it against either of the element references, since both
are optional and both match <a/>.  So, you cannot "uniquely attribute" the
element to a particle.  UPA is violated.

Note:  it's the original <sequence> that is not legal.  We can't really
ask whether the instance is valid, because there is no legal schema
against which to try (though clearly, if the schema were allowed, the
instance would validate...for two reasons!)

UPA is an interesting tradeoff.  It makes validators simpler, because you
don't have to go down two or more branches of the content model to try and
see whether your instance matches.  At least as important, it makes things
easier for tools that do things like binding schemas to Java, relational
databases, etc.  In the first schema above you can map any valid instance
to: a variable a, an array of up to 5 b's, and a c.  The sequence that
violates UPA is somewhat harder to map;  done in the obvious way, the
element <a/> could map to either of two variables, and crucially,
different processors might reasonably make different decisions.

I hope this helps.

Noah

[1] http://www.w3.org/TR/2004/PER-xmlschema-1-20040318/#Particle_details

--------------------------------------
Noah Mendelsohn
IBM Corporation
One Rogers Street
Cambridge, MA 02142
1-617-693-4036
--------------------------------------

Veerappan Saravanan-ESV007 <esv007@motorola.com>
Sent by: xmlschema-dev-request@w3.org
05/03/04 05:00 PM

To:     "'xmlschema-dev@w3.org'" <xmlschema-dev@w3.org>
cc:     (bcc: Noah Mendelsohn/Cambridge/IBM)