Re: XML Schema: DOs and DON'Ts

[Martin Gudgin <marting@develop.com>]

----- Original Message -----
From: "Kohsuke KAWAGUCHI" <kohsukekawaguchi@yahoo.com>
To: "Martin Gudgin" <marting@develop.com>; <xml-dev@lists.xml.org>
Sent: Wednesday, May 16, 2001 11:13 PM
Subject: Re: XML Schema: DOs and DON'Ts


>
> Wow, thank you for your comments!  I'm glad to know that at least one
> person has read my document.
>
>
> > 1.    I'm curious as to how you would validate an element that had
either
> > attributes and/or element descendants without using complex types. I may
> > have missed a major point here but I can't see how to validate;
>
> Well, as you know, completely removing the word "complexType" from your
> schema is impossible. What I wanted to say is one should use complexType
> only under <element> element, and one should never declare complexType
> globally.
>
> So as for your example, I would suggest
>
> <element name="person">
>   <complexType>
>     <group ref="person.contents"/>
>     <attributeGroup ref="person.attributes"/>
>   </complexType>
> </element>
>
>
> > without defining a complex type to describe the content of person. So,
are
> > you saying "don't use complex types" or are you saying "don't use named
> > complex types"?
>
> The latter is my intention. Don't use named complex types.

OK, that's fine I just wanted to make sure I understood your point.

>
>
>
> > 2.    Restriction gives you more that just error checking. You also get
a
> > type relationship between the base type and the derived type.
>
> Well, that is true. We get type hierarchy from complex types. But I
> wonder what do we get from type relationship. I appreciate if you would
> teach me what do we get from them.
>
> In the future, maybe we'll have tools to do something useful with those
> relationships. But there is no such tool. And tools like relaxer reveal
> that tools can be equally useful without explicit type hierarchy.
>
>
> Also, complex types give us type substitutability, which I failed to
> mention.
>
>
>
> > 3.    At the end of the day model groups only actually get used when
they
> > are referenced as part of a complex type definition.
>
> Umm, I couldn't understand your point.

Model groups on their own are useless. They only actually come into play in
a schema when they are referenced from inside a complex type definition. I
*thought* you were trying to say that model groups could be used on their
own to do validation. It came down to the same question as named vs
anonymous complex types in the end...


>
>
> > 4.    Regarding your comment about attribute groups, why not just use a
> > local attribute declaration? That *is* what you get anyway despite the
> > syntactic sugar that, it would seem, makes you think otherwise.
>
> Because I want to declare common attributes once and reference them
frequently.

OK, perfectly valid for that use case but that is *not* the use case you
give in your document. The reason you give for using attribute groups is
because you don't think people should use global attribute declarations.

>
>
>
> > <foo:person xmlns:foo="http://best.practice.com">
> >   <familyName> KAWAGUCHI </familyName>
> >   <lastName> Kohsuke </lastName>
> > </foo:person>
> >
> > is 'bad use of XML namespaces'. I have lots and lots of places where I
use
> > *exactly* that approach and it works very nicely for what I do. I don't
> > think you can call this one way or the other. Neither approach is
*wrong*,
> > they're just different
>
> I agree that there is no document that defines what is right and what is
> wrong in XML namespaces. So I wrote what I believe.
>
> I don't deny that my article contains some exaggerations :-)
> But that statement about XML namespace is no exaggeration nor
> simplification. It's my true feeling.
>
> I know some people uses namespaces in that way, but I never really
> understand why does this make sense. But I may miss the big point.
>
> Again I'd appreciate if you would tell me why you wrote
>
> <foo:person xmlns:foo="http://best.practice.com">
>   <familyName> KAWAGUCHI </familyName>
>   <lastName> Kohsuke </lastName>
> </foo:person>
>
> instead of
>
> <person xmlns="http://best.practice.com">
>   <familyName> KAWAGUCHI </familyName>
>   <lastName> Kohsuke </lastName>
> </person>

Perhaps because I'm mapping from the following Java type;

package best.practice.com;

public class person
{
    String familyName;
    String lastName;
}

I think a lot of people that are using XML for something other than document
markup may well end up going this way.

>
> And how you can put your fragment into *** of the following container?
>
> <?xml version="1.0"?>
> <root xmlns="uri:myMightyXMLContainer">
>     ***
> </root>

That's easy :-) *Always* put xmlns='' on any qualified element;

<p:person xmlns:p='http://best.practice.com' xmlns='' >
  <familyName>Gudgin</familyName>
  <givenName>Martin</givenName>
</p:person>

Default namespace declarations are evil! Putting xmlns='' in is defensive
"programming"

>
>
> > 7.    Chameleon schemas. I don't feel strongly about this but another
way of
> > looking at the problem is this; use local declarations instead if global
> > declarations.
>
> Apparently there are many things you can't write without global
> declarations. You can't use local declaration "instead of" global ones.
> You can only use local ones along with global ones.

Agreed. In general you need at least one global element decl unless you use
xsi:type in the instance.

>
> However, I also failed to mention that local elements give you a power
> to reuse the same tag name. That's another truth.
>
>
>
> > P.S. All your namespace URIs for schema have a type they should be
> > http://www.w3.org/2001/XMLSchema *not* http://www.w3.org/2001/XMSchema (
> > note missing 'L' character )
>
> Oops! thank you.

No problem. I find it ironic that I managed to spell typo incorrectly...

Regards

Martin