Re: case against text/* overstated

On Saturday, October 4, 2003, 12:40:01 AM, Dan wrote:


DC> This overstates the case:

DC> "any such transcoding will make the self-description false and will
DC> cause the document to be not well-formed."
DC> http://www.w3.org/TR/2003/WD-webarch-20031001/#no-text-xml

DC> counterexample:
DC> 	<doc/>
DC> transcode from us-ascii to utf-8 or iso8859-1.
DC> The result is still well-formed.

Until you do further processing, such as adding an attribute with an
e-acute. Though you still might get lucky - its the statistical
vagarity of the breakage that is the problem.

And, clearly, transcoding <doc/> between UTF-8 and UTF-16 *will* cause
non well formedness in all cases.

Perhaps "any such transcoding will make the self description false and
cause the document to be not well formed <i>unless you get real
lucky</i>"

DC> Overall, I think that section is awkward and doesn't
DC> make a very good case.

I think that section is stellar and makes a crucially important point.
An architecture that encourages non-well-formed XML on the server id
fragile, and I am strongly against any architecture that promotes it.




-- 
 Chris                            mailto:chris@w3.org