RE: [xml-dev] URIs harmful (was RE: [xml-dev] Article: Keeping pa ce

404 is what I expected, but you are right that it 
isn't reliable.  In other words, the address box reacts 
to any string.  urn:xx:yyy doesn't turn blue, though, 
so the HTML engine doesn't see a URL.  

I do count it as dereferencing if it attempts to 
find something and comes back with an error message. 
It makes the attempt.   

The point is to write the spec in such a way as the 
user doesn't get surprises.  Use of HTTP in the URI 
always produces a surprise.  Second, if the relationship 
is one to many (for any URI, there are multiple representations 
that may be returned, and these should be consistent) then 
the architecture can't specify other than return a consistent 
error, or better, return a RDDL or catalog entry.  That 
isn't different from what we have now, except the emphasis 
changes from "stupid user" to "lazy author".

len

-----Original Message-----
From: Thomas B. Passin [mailto:]

Sort of interesting to see how the different designers figured out what to
do, but I do not count any of these as dereferencing. Do you?  Nor the 404
you brought into prominence.