Re: Orphaned annotations

On Mon, 18 Mar 2002, Nick Kew wrote:

> That's why I suggested a method for measuring document change
> (or more precisely, a family of equivalence measures).

OK, to follow up to myself, let's spell that out.

An annotation can include its own date, and a document last-modified
date - more useful where available.  But this doesn't help to measure
which changes are significant.

The solution Valet uses to track changes is hashing on an ESIS
representation.  The most useful hash for testing the validity
of annotation pointers is that on document elements.  I'd suggest
that these be properly integrated into annotea:

* An annotea client should compute a hash when it makes an annotation
* The server stores the hash within the rdf:
	<a:elementhash>LWDBNLW/24DqY</a:elementhash>
* Queries on the database return a hash.
* The client rendering an annotation computes a hash on the document
  at the time of reading it.  If this differs from the hash associated
  with the annotation, the pointer is invalidated.
* Document management software can use this to prompt an administrator
  to deal with annotations, or can do so automatically if this is
  considered appropriate.  At this point, if annotations are kept
  their hashes can be updated.

A more sophisticated document management system might want to use
more hashes (Valet already does).

>
> > Second, how to change the status of annotations either manually or
> > automatically as part of the review process is an interesting problem and
> > we have been discussing about couple of approaches (in our future to do
> > list). If there are use cases, ideas, solutions etc. we are interested in
> > hearing about them.
>
> OK, I think you should separate the manual and automatic cases.  Doing
> it manually is just a case of a software tool maintaining dependency
> information and basic housekeeping.  I find the automated situation
> more interesting: for example, if a document has changed, I want to
> be able to detect which annotations are affected and should be
> archived off or flagged for human attention.  The person doing the
>
> [1] Yes I know XPointer does that too.  There are cases when you *can*
> meaningfully refer to a range; my point it that to try and do so in
> the presence of change is not sensible.
>
> --
> Nick Kew
>
> Site Valet - the mark of Quality on the Web.
> <URL:http://valet.webthing.com/>
>
>
>


-- 
Nick Kew

Site Valet - the mark of Quality on the Web.
<URL:http://valet.webthing.com/>

Received on Monday, 18 March 2002 16:47:12 UTC