Re: Are *relative* URIs as namespace nemes considered harmful?

At 08:03 PM 2000-05-17 +0100, David Carlisle wrote:
>Incidentally is there a canonical base URI for a file on my filesystem 
>(say /users/davidc/file.xml)?


For example, it is appropriate to overload a file with multiple access
methods, each generating a different BASE version for the file.  There is
no standard rule designating which of these is canonical.


When you open the document through the file system, your should in effect
save <file://> as the BASE appropriate to
the open image of the file.  But someone else may have ftp: access to the
same file.  Because the FTP and HTTP servers typically replicate the file
system hierarchy at least to the extent that subtrees are preserved,
relative URLs can work transparently across the disparate root segments
provided by the respective BASEs observed in the different access methods.

>If relative namespace URI are allowed, and namspace names are taken to
>be absolute, is it clear what is the namespace of xxx in
>file.xml if the content of the file is
><xxx xmlns="abc"/>
>Or will different systems make different choices?

Your designation of " in file.xml" is not sufficient precedent to make the
BASE and resolution of abc clear.  The scope that will be searched for the
object named ./abc will vary with the access method or path used initially
to access the file.  There are different de_facto expansions for "./"

This becomes excrutiatingly clear when you use an individual-file logical
link to insert a file in the webserver's namespace.  The peers of the file
could be totally different for web access to its logical location vs. file
access directly to its physical location.



Received on Wednesday, 17 May 2000 17:14:05 UTC