[rdfweb-dev] foaf weblog'd

Seth Russell seth at r...
Wed Sep 25 00:47:26 UTC 2002

Dan Brickley wrote:

> You pop up and plug semref whenever I mention rdfs:seeAlso! Can we just
> take your objection to this use of rdfs:seeAlso as read? Or perhaps you
> could send a review of http://www.w3.org/TR/rdf-schema/ to
> www-rdf-comments if you haven't already?

Do you really think this would make it into rdfs ? If you do, I 
certainly will write it up and submit it.

> We've discussed this several times in several fora. In the Web there are
> no hard and fast guarantees. In the old RDF groups we didn't define
> rdfs:seeAlso as always pointing to an RDF/XML document, since there are
> many slight variants of that scenario (content negotiation; alternate RDF
> syntaxes; mixed-namespace XML) that would make that too restrictive a
> rule. If people really care, they can use other markup to describe the
> remote resource as being of dc:format application/rdf+xml. It is possible
> we could refine the wording in the RDF Schema spec to make it clearer that
> the target of an rdfs:seeAlso property is 'more information' in the RDF
> sense, rather than in some looser, human-oriented sense. It is a tricky
> distinction to capture accurately though; suggestions (to
> www-rdf-comments) welcomed.

I think we need rdfs:seeAlso just as defined ... it doesn't specify 
what the target grammar is at all ... just says there is more 
information there about the subject. As we discussed before, {semref 
rdfs:subPropertyOf rdfs:seeAlso}. 

> In the meantime, rdfs:seeAlso is deployable, and (with other markup)
> adequate to help folk get experience with harvesting and processing a web
> of interlinked RDF documents.

Sure it's deployable .. and i suggest people go ahead and deply it. 
You are squeing your conceptions here to some kind of discovery 
spidering .. like your skutter ... but, Dan, that's not the primary 
application where we need semref. We need semref to tell a client 
where to put a gadget that the human user may click upon and reasonably 
expect a reliable result .... just like a hyperlink. These are just the 
cases where going to a subordinate node (that has the format specified) 
or dereferencing is not something that the application programmer is 
inclined to do, or in most cases (as in RSS 2.0) is even capable of 
doing. The other application is where a RDF process is logically 
following some flakey RDF that is being read in from the wild west of 
the Internet. Sure that RDF process can go dereference and guess how to 
parse the target, but that makes a very inefficient process .. with 
semref the process is much more efficient. 

> I suggest we focus on the definition of
> rdfs:seeAlso and on tools, and that encouraging folk to use an
> as-yet-undefined property 'semref' could distract from efforts to get some
> real world experience with such a distributed dataset...
> Hope this doesn't sound too grumbly, I'm just wary of there being no clear
> way of doing things. For the purposes of FOAF and other rdfweb-dev
> experiments, let's please use rdfs:seeAlso for now. I've made a similar
> case in the RSS working group too.

On the other hand, what harm would come if you just define semref in the 
foaf namespace? Let people use it if they choose, or not if they choose 
not. Spidering could follow both seeAlso and semref with very little 
extra code. What do we have to loose? 

Seth Russell

More information about the foaf-dev mailing list