more on Techniques for 2.1

As an example of an attribute which is clearly "for machines" and should
not be presented in literal text form exept as a last resort, consider any
attribute of type IDREF, such as the FOR attribute on LABEL elements in HTML4.

Here the format semantics is clear, the value of the attribute is a
reference to another element in the same document.  

In presenting this information to the user, the form "Reference to [element
summary]" is clearly supported by the definition of the format, and clearly
easier to understand than the text of the ID of the referenced element.  In
this case we can specify a formatting of a processed version of the
attribute text which is guaranteed to capture all the information in the
attribute.  But these cases are rare.

If the ID given does not match an element in the document, then the
document is broken and the text of the broken ID reference should be
provided to the user in lieu of a summary description of the referenced
element.

Machine-interpretable attributes have machine-generated improvements on the
text value of the attribute, as how to present the associated property to
the user.  The property still needs to be reported, but it may be amenable
to an automatic upgrade in understandability, as in this case.

This is the basic philosophy: the whole infoset as catalogued in the DOM is
the content of the document.  The user agent may or may not have access to
friendlier ways to display this content than just the text values in the
structure of the markup language.  The text values are the floor above
which one can go with improvements, but below which one should not fall.
And it should all be reachable somehow.  All those attributes are there to
encode information, regardless of whether the default view presents them as
text or expresses the information as properties on some of the rest of the
content.  

Al

Received on Thursday, 20 April 2000 17:23:43 UTC