What is the best practice for an XML schema describing an element with =
one or more
attributes plus data content, where the data content is not a supplied =
data type and
neither is it shared by other tags?

For example:

<tag1 attr="value1">some data</tag1>
<tag2 attr="value2">other data</tag2>

where I would like to describe each <tag> content something like

<xsd:restriction>
    <xsd:enumeration value="red"/>
    <xsd:enumeration value="blue"/>
    <xsd:enumeration value="green"/>
</xsd:restriction>

The short question is simply: what's the preferred way to do that?

The separate longer question is more specific.  In my particular case, =
there are many
different elements, each with a different description of the data =
content.  There are only
a small number of different attributes, where the same attribute could =
be used by several
elements.  It happens that the attribute definition is loosely related =
to the value's base type
(that is, the data content of all elements using a certain attribute =
might be based on
xsd:string, the data content of all elements using a certain other =
attribute might be based
on xsd:int, etc.)

That suggests to me that I should create my own base types, one for each =
attribute, that
describe both the attribute and the base type of the data content.  Then =
my element
definitions would be derived and further restrict the data content.  I =
can't figure out how
to do that.

So here's the exercise for the interested, what's the most efficient XML =
schema for
something like this, understanding that there are many more elements and =
maybe a few
more attributes:

element A accepts attribute "text" and string data "red", "blue", =
"green".
element B accepts attribute "text" and string data with a length less =
than 5.
element C accepts attribute "number" and numeric data between 100 and =
200.
element D accepts attribute "number" and numeric data between 1 and 10.

Thanks,
Lee