Re: It's about Reading -- not Writing...

[Danny Ayers <danny.ayers@xxxxxxxxx>]

On Wed, 18 Aug 2004 11:20:37 -0400, Bob Wyman <bob@xxxxxxxx> wrote:

>         RDF may contribute, as Dan Brickley says, "a strategy for
> decentralizing vocabulary design." However, the goal of the Atom effort is
> to contribute a "strategy and means for communicating information." The
> problems of designing vocabulary and of employing vocabulary (communicating)
> are distinct.

True, but one is dependent on the other.

>         The RDF camp seems enthralled by the glory of being able to easily
> *express* a vast range of ideas with what is, at its core, a very simple set
> of rules. However, we've learned over and over that RDF is more often
> written then read. Non-RDF XML is the language of readers -- not RDF. The
> reason is simple... RDF gives so much freedom to the writer that that reader
> has little hope of understanding what is written. 

RDF is a versatile model in which information can be expressed, when
passed from place to place communication can take place. But RDF/XML
is more tightly constrained than XML in general.

What one can understand
> about something written in RDF is often no more meaningful then the examples
> that Danny Ayers provides elsewhere. (e.g. Some unknown thing is said to
> have a property which has unknown, but named, properties and that something
> is said to have some undefined relationship to some other unknown thing...
> How is this useful?)

You can make comparisons between things without knowing everything
about them. You can make simple inferences to fill in the gaps, making
maximal use of whatever information is available. An approach such as
RDF inference may not be any more useful in a particular situation
than not using partial information, but discarding information can
never more useful than extracting what you do understand.

Do you know how you would get from your house to mine? Probably not,
but you know how to figure it out from the available information.

>         If one is communicating in a small group it is reasonable to utilize
> private vocabularies -- to produce "jargon" as it were. Small groups do this
> constantly -- with our without RDF. But, if you wish to communicate to a
> large group then, in order to *communicate*, you must express yourself with
> a broadly understood and more rigorous language. 

I agree with the point in the context of machine languages, but it's a
stretch to say English (for example) is rigorous.

The opportunity for variety
> and creativity in message construction is reduced as the size the audience
> increases.
>         The key challenge of Atom is, I believe, to enable speaking to large
> audiences. Thus, Atom should focus first on the problem of providing a well
> and rigorously defined vocabulary that can be shared efficiently by a large
> audience. Atom should be optimized for communications. I believe that this
> means that Atom should rely on rigorously defined XML -- complete with
> schemas, etc.

I agree. The problem is that if extensions are required, then there
needs to be some kind of commonality (rigour) beyond what
XML+namespaces alone offers.

>         RDF is certainly a useful tool for "vocabulary design" and can be
> very effective in small, closely-coupled groups -- thus, RDF can be useful
> to those who are experimenting with and designing early vocabularies that
> might later be formalized for broad use. In some small groups, the focus of
> communications is so limited that such formalization may not ever be
> justified or even necessary. 

This seems a strange way of looking at it - RDF can be just as useful
for working on web scales too.

Given this, I think Atom should provide a means
> to encapsulate RDF within its XML -- to enable experimentation and
> communication within fringe groups.
>         However, once a vocabulary has been "designed" as a result of small
> group RDF use, it simply doesn't make sense to continue using RDF. Once the
> vocabulary, or its core, has become fixed and well understood, the
> flexibility of RDF becomes both a burden (because of bulk, parsing
> difficulties, etc.) as well as unnecessary to communicate those things that
> have been formalized. Thus, *formal* extensions to Atom, intended to be used
> by the broad market, should be defined rigorously in as non-RDF XML.

So you are in effect proposing one single, growing, megalanguage (with
RDF as a kind of prototyping system). I'm not optimistic about that
approach to scaling.

>         Insisting on XML as the format for broad audience communication does
> nothing to limit the RDF people. As shown by GRDDL[1], you can convert
> easily between XML and RDF. The RDF-folk can view the whole world as RDF if
> they want... On the other hand, insisting on well and rigorously defined XML
> for Atom will make it much more likely that the users of Atom will be able
> to accomplish their communications goals -- because they will have agreed
> upon common vocabularies with some sense that their audience, *the readers*
> of Atom files, will understand.

I don't really disagree, having core Atom as non-RDF XML isn't really
much of an issue.

>         The Atom extensibility mechanisms should address at least two
> concerns:
>         1. Adding additional "standard" elements that are expected to be
> understood by *all* or a large number of Atom readers.
>         2. Providing the flexibility needed by vocabulary designers, small
> groups, and others who require or desire non-standard vocabularies.
>         The mechanisms which are most appropriate to handle the first
> concern are probably not the same as those which are most appropriate for
> the second.

Couldn't a scalable, modular mechanism cover both? 

>         Please remember: It's about reading -- not writing...

There has to be something to read, surely..?

Cheers,
Danny.