Boundaries of Atom (was Re: to/cc/bcc extension)

[Niklas Lindström <[e-mail]>]

Hi Alex,

On Sat, Oct 16, 2010 at 12:54 AM, Alexander Johannesen
<[e-mail]> wrote:

> Not meant to fixate this on the follow-up'er, but everything you ever
> wanted and extend Atom with is just conceptual things. There is no
> difference between a defined ontology with all the entities and
> relationships of Atom, and in fact you can define the whole shebang in
> any ontological language. It just so happens that Atom is a short-list
> of ontological expression defined in a format of sorts.

This is true. For generic data in a broader sense, RDF (linked data)
can be much more appropriate, in it being intrinsically open and
semantically precise. My own interest in Atom is due to its focus on
the specifics of capturing manifests for web (http) resources that
appear over time. I find that this simple, relatively closed scope,
tied to format (syntax), is what facilitates the ease of adoption for
many "mechanical" content syndication scenarios.

Whenever it's used for more than that I think it begins to "bulge",
and crosses over into domains where RDF is more appropriate to use
directly. Not for every extension of Atom of course (e.g. archives,
checksums, deletions and other "close to HTTP and REST" subjects), but
for connecting to domains outside of the scope of web resource
syndication.


> Where do the data model, format it's defined in and the ontology stop and start?

I believe that the Atom format, like many formats, represents the sole
formal structure of its underlying data model. This model can be
considered underdefined -- regarding e.g. what the atom:id identifies,
what the subject of the link relations is, etc. But that's probably
fine, since Atom does not purport to define these details, only to be
usable for certain data scenarios.

RDF on the other hand, has several serialization forms. There are no
"extension elements", just relations (identified by URI) between
subjects (URI:s) and objects (URI:s or literal values). This same
model is then used to define vocabularies (schemas/ontologies), by
using the relation URI:s as subjects and telling more about them.


> At some point the Atom model becomes so complex that it loses its easy
> applicability, and perhaps should enter the realm of ontological work.
> Anyone done any mappings of this so far?

AtomOwl is a common formulation of the Atom model in RDF:

    <http://bblfish.net/work/atom-owl/2006-06-06/AtomOwl.html>

SIOC is also close to the same domain, but focused on communities,
forums etc. in a more broader (richer) sense:

    <http://sioc-project.org/>

(SIOC is one of the core ontologies used to semantically representing
content in e.g. Drupal 7: <http://groups.drupal.org/node/44094>.)

There is also the OAI-ORE (Open Archives Initiative Object Reuse and
Exchange) which exemplifies the close relation between what extended
Atom and RDF can be used for:

    <http://www.openarchives.org/ore/>

Finally, there has been attempts to define how to extend Atom with RDF
more directly, by embedding RDF statements to "embellish" entries or
linked resources therein with rich, precise metadata. For example Atom
Triples and Yahoo:s DataRSS:

    <http://tools.ietf.org/html/draft-nottingham-atomtriples-00>
    <http://developer.yahoo.com/searchmonkey/smguide/datarss.html>

(There is work remaining in this area, if anyone is interested.)


Note that I don't make any claims here as to what is "best" for every
scenario. I only wish to highlight different concerns, scopes and
boundaries, and the different semantics and technologies used to
express and manifest those.

All the best,
Niklas