Re: Significant W3C Confusion over Namespace Meaning and Policy

Nicely said, Roy.

One comment ...

On Fri, Feb 11, 2005 at 12:17:32PM -0800, Roy T. Fielding wrote:
> I do not know of any
> justifiable reason why the set of names within a namespace
> cannot be extended without changing the definition of other
> names within that space, and thus without changing the
> distinctiveness of those names within their own namespace.

I can think of one reason; XML namespaces, for better or for worse(*)
are being used in place of media types by a significant number of Web

I know you're aware of this Roy, but for the benefit of others,
consider the message;

HTTP/1.1 200 Ok
Content-Type: application/xml
<html xmlns="">

I've observed[1] that most browsers (except IE), in the presence of a
*/xml media type, sniff out the namespace in order to determine the
intended semantics of the representation (aka, they "dispatch" on the
namespace).  As a result, the response message above would be considered
by these agents to have identical semantics to a message which included
the same XHTML document, but used the application/xhtml+xml media type.
IMO, this means that one needs to consider the implications of adding a
name to a namespace in the same way that one needs to consider minting
a new media type if a backwards incompatible revision to a document
format is produced.  Of course, without IE supporting this behaviour,
it's not yet pervasive, but I would expect that any future IE revision
would attempt to match this behaviour with other browsers.  The presence
of document formats without media types also encourages the generic
types be used too (e.g. XSLT 1.0, XForms).

Yet another new issue?


(*) IMO, worse, since (specification) layering is violated and
visibility reduced.  But it doesn't seem too harmful to Web
architecture, as it just limits expressibility.  For an example of
that, see;

Mark Baker.   Ottawa, Ontario, CANADA.

Received on Sunday, 13 February 2005 12:26:53 UTC