Ian Mayo wrote:
> I'm aspiring to use atom:category elements in a number of 
> schemes to define a series of attributes of my data items, as 
> espoused at:
> http://www.majordojo.com/2006/05/overloading-atomcategory.php
> 
> I'm relatively comfortable with the representation of privacy 
> and security via scheme, term, and label, but am stumbling at 
> the challenge of representing these concepts in a schema.  It 
> appears that XSD isn't able to define a set of valid terms 
> for one scheme, with another set that are only valid in 
> association with another scheme.
>
> Having had a look around at various specifications based on 
> Atom (including the Google data ones), they don't go as far 
> as defining a computer-readable schema, they stop at the 
> human-readable schema (as for the Atom spec).
> 
> Would somebody mind giving me an idea of why it appears to 
> have become acceptable to define these data-exchange 
> mechanisms just using a human-readable format that can't be 
> enforced by computer?  For a system involving the exchange of 
> data using an Atom-derived format, the lack of a 
> computer-enforceable schema seems a real handicap.

That is one reason why I recommend using atom:category only for user
categorization, and extension elements for everything else. If you don't
want users to add, remove, and change the categories then using
atom:category will create problems because (1) there is no standard way of
communicating what schemes/categories should be exposed to the end-user for
editing, (2) there is no (standard) way to communicate your categorization
restrictions (schema) to a client that allows the user to change the
categorization, (3) even if there was such a standard schema mechanism, it
is very difficult to convert that machine-readable schema into a user
interface that makes sense for the user.

That blog post linked to James's explanation of how his software uses
atom:category, but it seems to me that his software could have used
extension elements instead of categories for the application-specific
metadata without any trouble.

- Brian

Reply via email to