Re: Fielding on MOVE & COPY

>I don't mean just file attributes, but metadata in general -- author, title,
>publisher, subject, etc.  Most metadata stays the same whether a document is
>MOVEd or COPYd, but there are exceptions.  URN, if implemented, is another
>attribute that should never change when a document is MOVEd, whereas a COPY
>of a document should get its own URN.  Although work on metadata on the Web
>is very much "in progress" (URC, Z39.50, STARTS, etc.), I think it is an
>important part of document management that should be encompassed by the Web.
>It should be possible to set metadata using HTTP, however it turns out to be
>expressed.  And the metadata should get updated properly when a document
>gets COPYd or MOVEd.


I agree with the objectives but not with the conclusion. I see the 
keeping of metadata consistent to be a problem for the metadata system
and not for http.

The reason for this is that metadata can be external to the object.
The assertion "http://foobar.com/wellcome is a load of rubbish" is
not something that foobar.com are likely to want to store, manage or
otherwise be involved with. 

Whether a piece of metadata which relates to a resource which is bound
to name A should be in some way bound to the resource created at name
B through the operation COPY A->B is not a simple issue. In my view it is 
something that will have to depend on the particular assertions 
being made.


For example say I have a copy of the CNN site for 6th Sept 1996. I 
annotate this site with some political commentary "Operation Adios 
Bob Dole". In this case I am making an assertion relating to the
content of the message which is intended to be linked to the name
of the resource (ie the annotation has a secondness property wrt the
resource if we adopt Pierce's trichotemy).

Now say that someone comes along from the Dole camp and makes
an annotation "What do you expect from those pinko lefties from
Massachusetts?". In this case the annotation is clearly connected
to the site hosting the annotated copy of CNN and is not an
annotation on the CNN document.

I don't think that this type of issue can be settled except by 
considering the prperties of each assertion with respect to the 
individual operations. We might set out a set of laws which may
or may not be satisified by various assertion types.

IE "MOVE Invariance" :

      f(A)
-----------------
Move (A-B) ; f(b)

(if f(A) is asserted before we move the resource at A to B then
f(B) follows).

The assertion "Authored" clearly has move invariance, but "created"
does not since creation is usually interpreted as being the act
of binding a resource to a location.


		Phill

Received on Friday, 6 September 1996 18:06:37 UTC