[
Lists Home |
Date Index |
Thread Index
]
Okay... so some people think Namespaces should become an integral part of
XML. Some think that it should not. Some see no use for Namespaces, at
least as far as their own XML involvement goes. Others see Namespaces as
useful, but problematic. So... how about if we provide a alternative to
Namespaces? After all, it is sometimes better to start from scratch than to
Here is one idea...
The way I view Namespaces is that it is a method by which one can mix
multiple schemas that contain non-exclusive vocabularies. A namespace value
can also be used as a method to identify a schema that a document conforms
to. As a result, the concept of namespaces is tied more to schema
definitions than it is to XML itself. So how about working Namespaces into
schema languages instead of into XML?
For instance, suppose you have the following two schemas (using a fictional
language):
<!-- URI: -//SchemaABC -->
<schema>
<element name="ABC"/>
<element name="Body"/>
<element name="From"/>
</schema>
<!-- URI: -//SchemaXYX -->
<schema>
<element name="XYZ"/>
<element name="From"/>
<element name="To"/>
</schema>
Now, suppose that you wanted to have a document that used some of each of
the schemas. In order to do this with Namespaces, you might name each
schema into a separate namespace something like the following:
<abc:ABC xmlns:abc="-//SchemaABC" xmlns:xyz="-//SchemaXYZ">
<abc:Body></abc:Body>
<xyz:From></xyz:From>
<xyz:To></xyz:To>
</abc:ABC>
On the other hand, suppose instead that you were to create a third schema
that looked like the following:
<!-- URI: -//SchemaABCXYZ -->
<schema>
<include URI="-//SchemaABC" preface="abc:"/>
<include URI="-//SchemaXYX" preface="xyz:"/>
</schema>
Then an XML document might name the schema as follows:
<abc:ABC xmlns="-//SchemaABCXYZ">
<abc:Body></abc:Body>
<xyz:From></xyz:From>
<xyz:To></xyz:To>
</abc:ABC>
So far, what we have effectively done is grouped multiple schemas into a
single schema and thereby reducing the namespace to one at the document
level.
Now, a more robust implementation of a schema might look more like the
following:
<!-- URI: -//SchemaABCXYZ -->
<schema>
<include URI="-//SchemaABC">
<exclude-element name="From"/>
<alias-element name="ABC" alias="myABC"/>
<!-- Body element is being used as-is. -->
</include>
<include URI="-//SchemaXYZ">
<exclude-element name="XYZ"/>
<alias-element name="From" alias="Sender"/>
<alias-element name="To" alias="Recipient"/>
</include>
<element name="Timestamp">
</schema>
With this, the document might look like the following:
<?xml version="1.1" schema="-//SchemaABCXYZ"?>
<myABC>
<Sender></Sender>
<Recipient></Recipient>
<Body></Body>
<Timestamp/>
</myABC>
This method does not break DTDs (as far as I can see). It also gets rid of
the need for the XML Namespace mechanism. If someone still wanted the
namespace-style names, they could alias any that they want to use. On the
other hand, I can foresee some setbacks as well. I also realize that the
above examples are not necessarily the exact way things should be done or
that the above concepts are all-inclusive of what is needed within such a
concept to make this work. This would be an opportunity for the community
to come up with something (assuming the idea even has enough merit to start
with).
Some questions I can think of are:
1) How easy or difficult would it be to add this concept to existing schema
languages?
2) Could a common XML syntax (as shown in the last example) be used
regardless of the actual schema language used?
3) What would this mean from a performance point-of-view?
Well, let's hear it. Is this workable? Or is this just my total
misunderstanding of the problem?
---
Seairth Jacobs
seairth@seairth.com
|
