This conflict reflects the underlying conflict in the purpose of HTML.
While it supposedly describes the structure of a document, it's most
often used to tell a browser how to present a document. I believe that
we're going to keep seeing this conflict until some some means of
unifying the two distinct purposes of HTML is reached.
It was suggested that, in the grand scheme of SGML, HTML is correctly
viewed aas a presentation language that you translate documents in
other SGML DTDs into for display purposes. This seems reasonable, so
long as you keep the goal of being able to lose only presentation
information as you move from platform to platform in mind.
This means that new presentation tags need to be added with some care
as to how they are going to be represented on other browsers, as well
as some care as to how browsers that don't implement them are going to
behave when they see them.
IMG is a good example of ignoring the first criteria - the initial
implementation had no way to replace the missing information, and the
design was such that only a very limiting kludge could solve the
problem. IMAGE show how it's done right - you can bring all the power
of HTML to your aid in constructing the replacement information, and
there's a reminder that you need to do so when you write the thing. It
also has the right property when it's ignored by a browser - users get
the presumably lower-quality information.
CENTER is a good example of ignoring the second criteria. Because it's
a paragraph tag (at least, the only documentation available says it
can be used as one), ignoring it means that you lose paragraph
information. There has been a proposal that it be accepted with the
restriction that it can't directly contain text, which solves the
problem. Except that, unless popular browsers do something very ugly
when that restriction is violated, people will ignore it, and we'll