Ian Hickson wrote:
On Tue, 19 Jul 2005, Olav Junker Kjær wrote:
But the notion of conformance is still quite useful to authors and
authoring tools. E.g. if a META-element without any attributes appears
in a document, its clearly due to an oversight or a bug in some tool, so
it would be useful to have a conformance checker or authoring tool flag
this, even if a browsers will handle it somewhat gracefully (by ignoring
it).
Agreed.
I agree that we need to make conformance checking useful, of course. I
disagree that a blank <meta/> is necessarily a problem. Maybe the author
wanted to add some attributes dynamically later. Maybe he wants the DOM of
all his pages to be equivalent and at that point in his pages there simply
is no metadata to give.
...
The difficulty is in walking the fine line between useful and
over-constrained. For example, the fact that <ol></ol> is invalid in HTML4
is a real problem.
Agreed with the last paragraph.
One way of drawing the line might be, does dropping this requirement
result in a semantically-meaningful representation? An empty list
represents an empty list. But a <meta> without a 'name', or a <link>
without a 'href': these, per spec, represent nothing. They do not even
provide any structural semantics as <div> and <span> do; the document
has the same semantics as if the element did not exist.
~fantasai