Annotation protocols [was: How namespace names might be used]

-----Original Message-----
From: Tim Bray <tbray@textuality.com>
To: xml-uri@w3.org <xml-uri@w3.org>
Date: Saturday, June 10, 2000 6:29 PM
Subject: How namespace names might be used


>Namespace names either have URI syntax or are URIs.  Whichever is the case,
>how might we envision them being used in an ideal future?
>
>Level 1 idea: use the URI to retrieve the semantic resource.  I think this
>is more or less what TimBL and Dan are in favor of.  I'm on the record as
>thinking that this approach is pretty limiting, mostly because I believe
>that semantic resources come in lots of complex parts, including human-
>readable documentation, RDF, schemas, Java classes, and lots of other
>stuff that hasn't been invented yet.
>
>Level 2 idea: use the URI to start down a retrieval trail.  This is the
>idea, which I and others have talked up, of there being some sort of
>universal related-resource-clustering vocabulary - the word "packaging"
>has been used - if such a thing existed, and were conventionally placed
>where a namespace name points, this might be a real step forward for
>everyone.


Once you use a URI in XML, then all these questions immediately
fall on the other side of the wall - the URI side, not the XML side.
(Discusion on http://lists.w3.org/Archives/Public/uri/ in theory)

This is a really important part of the modular deisgn.  It is so many
specs have been able to be produced without having to rechart
or worse reinvent the URI system.  The URI identifies.
Many many things can as a result be done, but one doesn't
have to worry about what in the namespace spec.

While it is interesting to wonder about the languages which might
be available to describe a namespace, and it is might be useful
to have exacmples for input to thought experiments,

>But... the more I think about the packaging idea, the more it seems
>insufficiently flexible and general.  At the end of the day, it seems
>like all the different kinds of related resources (stylesheets, type
>definitions, procedural code, schemas) ought to somehow become active,
>and respond to call-by-name.  I.e. there ought to be a way to broadcast
>an appeal for stylesheets that can handle vocabularies named by
>http://a.b.com/ns37, or Java classes that can generate audio output
>from vocabularies named http://a.b.com/ns39; this is a many-to-many
>mapping we're talking about here, because a stylesheet resource could
>probably "know about" a wide variety of vocabularies (e.g., DocBook
>derivatives) that it's capable of handling.


There is a commonly asked for functionality - to ask one
resource what it thinks about another resource.  It is
often called the annotation server functionality.
It is more complicated than the dereferencing case,
in which you look for information which is definitive,
because it takes two parameters - the URI of the resource
you are interested information about and also that
of the service you are querying.

Such a protocol was standardized for the PICS
technology but that was pre-XML. Netscape's
"what's related"  service does this using
a HTTP query to an annotation service to return RDF.
There is an annotation project which Ralph Swick (w3C)
et al from W3C presented on devday at WWW9.
I think it is fair to say that the common form is
simply to use HTTP and the query syntax,
when querying service X for stuff about Y
as X?url=Y where Y has to be encoded of course.

There may be W3C standardization activity
coming out of this work if there is sufficient
interest.

The think which makes this project interesting is that
it is I think the first which allows the user to
an easy usre interface to juggling which annotation
servers he or she is using at any one time.
This is different from the crit.org, thirdvoice.com
etc models in which you only work with one "authority"
at a time.  These can work by produceing a single
annotated mirror of the entire web - and so don't
need annotation protocols.


>Are any of the existing Internet protocols a candidate for this
>kind of lookup-by-name?  I don't think content-negotiation goes nearly
>far enough.  Pardon me for blue-skying it.  -Tim [Bray]

yes, HTTP

Content negotiation does not of course do this at all.
It is part of a two-URI protocol, not a three-URI protocol.
It applies just as well of course to a dereferncing or
to an annotation query, in that either

So these are all interesting things one can do with URIs.
I am focusssing on getting XML to use a URIs
so DOM and everything else can move on.

Tim

Received on Monday, 12 June 2000 13:22:52 UTC